Hey all, I run an AI tools directory and thought I’d take the time to share some of my top picks for GPT-powered tools that create visual presentations/slides. Keep in mind none of these will completely replace manual work if you want something very high quality, but they do get the job done and takes out 90% of the work required. Without further ado, here’s a few that I’ve tried and liked, as well as my thoughts on them:
Plus AI for Google Slides- Great for Work; Presentations with Live Data in Snapshots
A fantastic tool for automating and enhancing my Google Slides presentations. Plus AI lets you start with a brief description of the presentation you need an an AI-generated outline is created, which you can then adjust according to your requirements. In addition, it lets you make ‘Snapshots’ from any web content which can be embedded and updated in my slides or documents with just one click. This is particularly useful for my team meetings and project reports as it significantly reduces preparation time. It’s available for free on the Google Marketplace as an add-on for GSlides.
Tome – Great for Business Storytelling
Generates a narrative based on a simple prompt, turning it into a presentation, outline, or story with both text and images. I found it very efficient for creating dynamic, responsive presentations, and appreciated how the AI could automatically cite sources or translate content into other languages. It’s an intuitive tool for anyone who needs to deliver compelling stories or presentations, from founders and executives to educators. A standout feature is the ability to embed live interactive content, such as product mockups and data, directly onto your page, bringing the storytelling experience to life. It’s available for free as a web app, with integrations for apps such as Figma, YouTube, Twitter, and GSheets.
STORYD – Business Storytelling, with Script Generator
This tool has truly revolutionized my approach to data presentations. By simply providing a brief summary of my topic, StoryD employs AI to script, design, and generate a presentation in less than a minute. Not only does this tool save me an immense amount of time, but its built-in ‘storytelling structure’ enhances the communicability and impact of my data. I also appreciate its customization options, such as themes, fonts, colors, and a plethora of layout options. The free limited beta version offers enough for the casual user, but the pro version at $18/mo adds useful features like team collaboration and real-time editing. Available as a web app.
beautiful.ai – Great for Visually Appealing Slides
A considerable time saver for anyone frequently creating presentations. Beautiful.ai provides a broad collection of smart slide templates, enabling you to build appealing and meaningful presentations swiftly. I was particularly impressed with its ability to automatically organize and design content in minutes, irrespective of your graphic design experience. It also offers slide templates for various needs, from timelines, sales funnels, SWOT analysis, to more specific ones like data & charts, visual impact slides, and so forth. The free trial is more than adequate for getting a feel of the service, and their paid plans start at $12/mo. It’s available as a web app and integrates with cloud platforms (i.e. Dropbox and Google Drive).
Albus – Knowledge Presentations/Cards/Map
Changes the way you typically interact with knowledge and facts; it harnesses the power of GPT to create an engaging and exploratory learning experience around any topic. Basically you start with with a single question and prompt, and you get a fact card, which you can then expand into other cards and images. I appreciate the way it opens up new perspectives and angles, allowing me to dive into a subject, ask questions, and organically grow my understanding. The ability to add notes and images to organize my board further enriches the experience. And when it’s time to share, I love how Albus AI facilitates controlled content presentation. With Albus AI, it’s not just about learning, but also about the journey of discovery. It’s available as a web app, and currently in Beta.
Decktopus – Great Overall for Work/Business, “Microsites”
Decktopus AI takes the pain out of crafting presentations. Simply key in a topic and it generates a fully fleshed out deck in an instant, which is a boon for my quick-turnaround needs. Its one-click design feature and auto-adjusted layouts simplify the customization process, saving me the headache of manual tweaking. I also appreciate the built-in tools such as image & icon suggestions, tailored slide notes, and extra content generation which further streamline the creation process. Its additional features, like voice recording and real-time audience feedback collection, elevate my presentations to a new level. For quick, professional-looking presentations, Decktopus AI is my go-to. It can also handle generating micro-sites (basically something that’s between a LinkTree and an landing page in terms of complexity). It’s available as a web app for free.
Gamma – Good Alternative to Decktopus
A fresh take on presentations, Gamma marries the depth of documents with the visual appeal of slides, powered by AI for efficiency. It lets me draft ideas quickly and the AI transforms them into professional-looking presentations in a snap. The interface is incredibly intuitive, allowing for nested cards for detailing and the ability to embed various forms of content, including GIFs, videos, charts, and websites. My favorite feature is the one-click restyle, removing the tedious task of manual formatting. Sharing the content is simple and works on all devices. Plus, it offers built-in analytics, which adds a nice touch to understand audience engagement.
SlidesAI – Text to Slides for Google Slides
A real game-changer for those frequently tasked with creating presentations. SlidesAI integrates seamlessly into Google Slides, transforming your raw text into professionally-styled slides in just seconds. The AI parses your input, breaking it down into digestible, summarized points, even providing automatic subtitles for each page – all in over 100 different languages. It not only eliminates the tedious task of manual slide creation but also offers a real time-saver when summarizing longer texts. While the free tier allows you to create up to 3 presentations a month, I personally found value in the Pro plan for the higher character limit and additional presentations per month. It’s available on Google Marketplace as a add-on for Slides.
MagicSlides – Idea to Slides for Google Slides
Transforms ideas into professional-looking Google Slides in seconds, eliminating the tedious work of designing and creating slides from scratch. I just input the topic and slide count, and it auto-generates a presentation for me, complete with relevant images and eye-catching layouts. Personalization options like theme selection, font choice, and color palette further enhance the final result. The app also supports over 100 languages, which is immensely helpful when I’m dealing with international projects. Like SlidesAI you get 3 free presentations per month and it’s available as an add-on for Google Slides.
Feel free to try these tools for yourself to get a feel for GPT’s potential as a visulization tool, and I hope this list will help at least some of you save some time for your next presentation at work!
ChatGPT explains (in morbid detail) what would happen to a man’s body if he was in a submarine at Titanic depth while it imploded
A submersible would be under immense pressure at the depths of the Titanic wreckage, which lies about 2.37 miles (3.81 kilometers) below the surface. At this depth, the pressure is over 370 times atmospheric pressure, or about 5,500 pounds per square inch (psi).
If the submersible were to suddenly implode, the effect on the human body inside would be catastrophic. The sudden compression of the environment around the man would almost instantaneously crush his body. This wouldn’t be a gradual process; it would happen in less than a second due to the enormous and immediate change in pressure.
The body would be subjected to rapid compression, causing immediate and severe trauma. Essentially, every part of the body that contains gas, including the lungs and the gastrointestinal tract, would be crushed or imploded. The water pressure would also force water into body cavities such as the nose, mouth, and ears. This rapid compression could cause severe internal injuries, including hemorrhage and organ damage.
Due to the speed of the implosion, it’s unlikely the individual would experience much, if any, pain. Unconsciousness would likely occur almost instantaneously due to the severe trauma and lack of oxygen.
In terms of visual details, the implosion would cause an immense shockwave in the water, creating a sudden cloud of debris consisting of the destroyed submersible and, unfortunately, the remains of the occupant. Following the implosion, the water would rapidly rush back into the void, contributing further to the turbulent scene.
It’s worth noting that these circumstances are hypothetical and based on current understanding of deep sea pressure and its effects on the human body. In reality, safety measures and design standards for submersibles aim to prevent such catastrophic failures from ever occurring.
Boost your advertising and social media game with AdCreative.ai – the ultimate Artificial Intelligence solution. Say goodbye to hours of creative work and hello to the high-converting ad and social media posts generated in mere seconds. Maximize your success and minimize your effort with AdCreative.ai today.
OpenAI’s DALLE 2 is a cutting-edge AI art generator that creates unique and creative visuals from a single text input. Its AI model was trained on a huge dataset of images and textual descriptions to produce detailed and visually attractive images in response to written requests. Startups can use DALLE 2 to create images in advertisements and on their websites and social media pages. Businesses can save time and money by not manually sourcing or creating graphics from the start, thanks to this method of generating different images from text.
Using artificial intelligence, Otter.AI empowers users with real-time transcriptions of meeting notes that are shareable, searchable, accessible, and secure. Get a meeting assistant that records audio, writes notes, automatically captures slides, and generates summaries.
Notion is aiming to increase its user base through the utilization of its advanced AI technology. Their latest feature, Notion AI, is a robust generative AI tool that assists users with tasks like note summarization, identifying action items in meetings, and creating and modifying text. Notion AI streamlines workflows by automating tedious tasks, providing suggestions, and templates to users, ultimately simplifying and improving the user experience.
Motion is a clever tool that uses AI to create daily schedules that account for your meetings, tasks, and projects. Say goodbye to the hassle of planning and hello to a more productive life.
With its outstanding content production features, Jasper, an advanced AI content generator, is making waves in the creative industry. Jasper, considered the best in its area, aids new businesses in producing high-quality content across multiple media with minimal time and effort investment. The tool’s efficiency stems from recognizing human writing patterns, which facilitates groups’ rapid production of interesting content. To stay ahead of the curve, entrepreneurs may use Jasper as an AI-powered companion to help them write better copy for landing pages and product descriptions and more intriguing and engaging social media posts.
Lavender, a real-time AI Email Coach, is widely regarded as a game-changer in the sales industry, helping thousands of SDRs, AEs, and managers improve their email response rates and productivity. Competitive sales environments make effective communication skills crucial to success. Startups may capitalize on the competition by using Lavender to boost their email response rate and forge deeper relationships with prospective customers.
Speak is a speech-to-text software driven by artificial intelligence that makes it simple for academics and marketers to transform linguistic data into useful insights without custom programming. Startups can acquire an edge and strengthen customer relationships by transcribing user interviews, sales conversations, and product reviews. In addition, they can examine rivals’ material to spot trends in keywords and topics and use this information to their advantage. In addition, marketing groups can utilize speech-to-text transcription to make videos and audio recordings more accessible and generate written material that is search engine optimization (SEO) friendly and can be used in various contexts.
Recently, GitHub released an AI tool called GitHub Copilot, which can translate natural language questions into code recommendations in dozens of languages. This artificial intelligence (AI) tool was trained on billions of lines of code using OpenAI Codex to detect patterns in the code and make real-time, in-editor suggestions of code that implement full functionalities. A startup’s code quality, issue fixes, and feature deliveries can all benefit greatly from using GitHub Copilot. Moreover, GitHub Copilot enables developers to be more productive and efficient by handling the mundane aspects of coding so that they can concentrate on the bigger picture.
For faster hiring across all industries and geographies, businesses can turn to Olivia, a conversational recruiting tool developed by Paradox. This AI-powered conversational interface may be used for candidate screening, FAQs, interview scheduling, and new hire onboarding. With Olivia, entrepreneurs may locate qualified people for even the most technical positions and reclaim the hours spent on administrative activities.
Lumen5 is a marketing team-focused video production platform that allows for developing high-quality videos with zero technical requirements. Lumen5 uses Machine Learning to automate video editing, allowing users to quickly and easily produce high-quality videos. Startups can quickly and easily create high-quality films for social media, advertising, and thought leadership with the help of the platform’s built-in media library, which provides access to millions of stock footage, photographs, and music tracks. In addition, AI can help firms swiftly convert blog entries to videos or Zoom recordings into interesting snippets for other marketing channels.
Spellbook is an artificial intelligence (AI) tool that leverages OpenAI’s GPT-3 to review and recommend language for your contracts without you having to leave the comfort of a Word document. It was trained on billions of lines of legal text. This artificial intelligence tool can be used by startups in drafting and reviewing agreements and external contracts to identify aggressive words, list missing clauses and definitions, and red flag flags. Spellbook can also generate new clauses and recommend common topics of negotiation based on the agreement’s context.
Grammarly is an AI-powered writing app that flags and corrects grammar errors as you type. A machine learning algorithm trained on a massive dataset of documents containing known faults drives the system. Enter your content (or copy and paste it) into Grammarly, and the program will check it for mistakes. Furthermore, the program “reads” the mood of your work and makes suggestions accordingly. You can choose to consider the recommendations or not. As an AI tool, Grammarly automates a process that previously required human intervention (in this case, proofreading). Use an AI writing checker like Grammarly, and you’ll save yourself a ton of time.
Chatbots are one of the most well-known uses of artificial intelligence. Computer programs called “chatbots” attempt to pass as humans in online conversations. They process user input using NLP algorithms that enable them to respond appropriately. From assisting customers to promoting products, chatbots have many potential applications. Chatbots on websites and mobile apps have increased in recent years to provide constant help to customers. Whether answering basic questions or solving complex problems, chatbots are up to the challenge. In addition, businesses can use them to make suggestions to customers, such as offering related items or services.
Keeping track of customer support inquiries can take time and effort, especially for smaller organizations. Zendesk is an artificial intelligence (AI)-powered platform for managing customer assistance. Zendesk goes above and beyond the capabilities of chatbots by discovering trends and patterns in customer service inquiries. Useful metrics are automatically gathered, such as typical response times and most often encountered issues. It also finds the most popular articles in your knowledge base so you can prioritize linking to them. An intuitive dashboard displays all this information for a bird’s-eye view of your customer service.
Timely is an AI-powered calendar app that will revolutionize how you schedule your day. It integrates with your regular software to make tracking time easier for your business. Track your team’s efficiency, identify time-consuming tasks, and understand how your company spends its resources. Timely is a fantastic tool for increasing the effectiveness and efficiency of your team. You can see how your staff spends their time in real-time and adjust workflows accordingly.
If you own an online store, you understand the ongoing threat of fraud. Companies lose billions of dollars annually to credit card fraud, which can also hurt your reputation. Through the analysis of client behavior patterns, fraud can be prevented with the help of AI. Machine learning algorithms are used by businesses like aiReflex to sift through client data in search of signs of fraud. It would be impractical and time-consuming to inspect every transaction manually. However, this can be automated with the help of AI, which will keep an eye on all of your financial dealings and flag anything that looks fishy. Your company will be safe from fraudulent activity if you take this precaution.
Murf is an artificial intelligence–powered text-to-speech tool. It has a wide range of applications, from speech generation for corporate training to use in audiobook and podcast production. It is a highly flexible tool that may also be used for voiceovers in promotional videos or infomercials. Murf is a wonderful option if you need to generate a speech but don’t have the funds to hire a professional voice actor. Choosing a realistic-sounding voice from their more than 120 options in 20 languages is easy. Their studio is easy to use, and you may incorporate audio, video, and still photographs into your production. As a bonus, you have complete command over the rate, pitch, and intonation of your recording, allowing you to mimic the performance of a trained voice actor.
OpenAI’s ChatGPT is a massive language model built on the GPT-3.5 framework. It can produce logical and appropriate answers to various inquiries because it has been trained on large text data. Because ChatGPT can automate customer care and support, it has helped startups provide 24/7 help without hiring a huge customer service department. For instance, the Indian food delivery firm Swiggy has used ChatGPT to enhance customer service and shorten response times, resulting in happier and more loyal customers.
Google’s Bard uses the Language Model for Dialogue Applications (LaMDA) as an artificially intelligent chatbot and content-generating tool. Its sophisticated communication abilities have been of great use to new businesses. New companies have used Bard to improve their software development, content creation, and customer service. For example, virtual assistant startup Robin AI has implemented Bard to boost customer service and answer quality. Startups can now provide more tailored and interesting user experiences because of Bard’s intelligent and context-aware dialogue production, increasing customer satisfaction and revenue.
Small business owners and founders often need persuasive presentations to win over investors and new clientele. Create great presentations without spending hours in PowerPoint or Slides by using Beautiful.ai. The software will automatically generate engaging slides from the data you provide, like text and graphics. Over 60 editable slide templates and multiple presentation layouts are available on Beautiful.ai. Try it out and see if it helps you make a better impression.
If you want to reach millennials and other young people with short attention spans, you need to have a presence on TikTok and Instagram. Dumme is a useful tool for extracting key moments from longer videos and podcasts to make shorts (short videos to share on social media). You may use Dumme to pick the best moments from any video or audio you post to use them in short. It will automatically create a short video with a title, description, and captions suitable for sharing online. Making a short video for sharing on social media can be done without spending hours in front of a computer.
The Open AI-backed firm Cohere Generate created the language AI platform. It helps organizations and startups save time and effort in creating large-scale, personalized text content. It employs NLP and machine learning algorithms to develop content that fits with the brand’s voice and tone. Use this tool to boost your startup’s online visibility, expand your reach, and strengthen your content marketing strategy.
Synthesia is a cutting-edge video synthesis platform that has been a huge boon to the video production efforts of new businesses. It uses artificial intelligence to eliminate the need for costly and time-consuming video shoots by fusing a human performer’s facial emotions and lip movements with the audio. To improve their advertising campaigns, product presentations, and customer onboarding procedures, startups may use Synthesia to create tailored video content at scale. For instance, entrepreneurs can produce multilingual, locally adapted videos or dynamic video ads with little to no more work. Synthesia gives young companies the tools to reach more people at a lower cost per unit while still delivering high-quality content.
Google has developed an AI-based service to combat money laundering. It has been trialed by HSBC to detect suspicious financial transactions. The aim is to mitigate one of the most challenging and costly issues in the financial sector: money laundering.
-
Money laundering is linked to criminal activities like drug trafficking, human trafficking, and terrorist financing.
-
This issue requires substantial resources and cross-state collaboration to track illicit funds.
-
Google’s AI-powered Anti Money Laundering (AML AI) service can analyze billions of records to spot trends and signs of financial crime.
Google’s AI Approach: The conventional methods of monitoring involve manually defined rules, which often lead to high alert rates but low accuracy. Google’s AI tool provides a more efficient solution. Google Cloud’s new AI-driven tool, Anti Money Laundering AI, eliminates rules-based inputs, reducing false positives and increasing efficiency in identifying potential financial risks.
-
Current monitoring products depend on manual rules, resulting in many false alerts and limited accuracy.
-
Human-defined rules are also easy for criminals to understand and circumvent.
-
The AI tool minimizes false positives, saving time, and enabling focus on truly suspicious activities.
Risk Score for Money Laundering: The AML AI tool creates a consolidated risk score, which is a more efficient alternative to the conventional rule-based alert system.
-
Instead of triggering alerts based on pre-set conditions, the AI tool monitors trends and behaviors.
-
The risk score is calculated based on bank data, including patterns, network behavior, and customer information.
-
This approach allows the tool to adapt quickly to changes and focus on high-risk customers.
Performance of the AI Tool: HSBC, as a test customer, found that the AI tool outperforms existing systems in detecting financial crime risk.
-
HSBC reported a 2-4 times increase in accurate risk detection and a 60% decrease in alert volumes.
-
This has helped reduce operating costs and expedite detection processes.
-
Google Cloud’s AML AI has enhanced HSBC’s anti-money laundering detection capabilities.
Source (TechMonitor)
Researchers from Yamagata University and IBM Japan have used a deep learning artificial intelligence model to discover four new geoglyphs in the Nazca desert of Peru, dating back to between 500 BC and 500 AD. The AI system accelerated the identification of these geoglyphs, making the process 21 times faster than human analysis alone.
AI Discovery of Geoglyphs: The team from Yamagata University and IBM Japan used a deep learning AI model to find new geoglyphs in Peru.
-
Geoglyphs are earth carvings that form shapes and lines, some of which can reach up to 1,200 feet long.
-
Four new geoglyphs were identified, depicting a humanoid figure, a fish, a bird, and a pair of legs.
Academic Debate: There’s disagreement among scholars about why these geoglyphs were created.
-
Some believe they were made to honor deities thought to observe from above.
-
Others suggest extraterrestrial involvement, with the lines serving as airfields for alien spacecraft.
Use of AI in Archaeological Research: Previously, finding new geoglyphs required researchers to manually examine aerial photographs, a time-consuming and challenging task.
-
The scientists trained a deep learning system to identify potential Nazca Lines based on previously found geoglyphs.
-
The AI system significantly sped up the process, working 21 times faster than human analysis alone.
Future of AI in Archaeology: Following this success, the Yamagata researchers plan to team up with the IBM T. J. Watson Research Center to extend their research to the entire region where the lines were discovered.
-
There are also plans to work with Peru’s Ministry of Culture to protect the newly discovered geoglyphs.
-
The team predicts that recent technological advances in drones, robotics, LiDAR, Big Data, and artificial intelligence will propel the next wave of archaeological discoveries.
Previous AI Contributions to Archaeology: Artificial intelligence has been previously used to solve other archaeological mysteries.
-
AI systems have helped identify patterns on land using satellite and sonar images, leading to discoveries like a Mesopotamian burial site and shipwrecks.
-
AI has also aided in translating ancient texts, as researchers at the University of Chicago trained a system to translate ancient inscriptions with an 80% accuracy.
- Researchers from Yamagata University and IBM Japan used a deep learning AI model to discover four new geoglyphs in the Nazca desert in Peru, which are ancient large-scale artworks etched into the earth.
- The AI system analyzed aerial photos of the area, working 21 times faster than human analysis, identifying potential Nazca Lines based on previously found geoglyphs.
- Following the success of AI integration in archeological research, the team plans to expand their research to the entire region, highlighting that AI technology may drive the future of archeological discoveries.
AI like ChatGPT, once known for providing detailed instructions on dangerous activities, are being reevaluated after a study showed these systems could potentially be manipulated into suggesting harmful biological weaponry methods.
Concerns About AI Providing Dangerous Information: The initial concerns stem from a study at MIT. Here, groups of undergraduates with no biology background were able to get AI systems to suggest methods for creating biological weapons. The chatbots suggested potential pandemic pathogens, their creation methods, and even where to order DNA for such a process. While constructing such weapons requires significant skill and knowledge, the easy accessibility of this information is concerning.
-
The AI systems were initially created to provide information and detailed supportive coaching.
-
However, there are potential dangers when these AI systems provide guidance on harmful activities.
-
This issue brings up the question of whether ‘security through obscurity’ is a sustainable method for preventing atrocities in a future where information access is becoming easier.
Controlling Information in an AI World: Addressing this problem can be approached from two angles. Firstly, it should be more difficult for AI systems to give detailed instructions on building bioweapons. Secondly, the security flaws that AI systems inadvertently revealed, such as certain DNA synthesis companies not screening orders, should be addressed.
-
All DNA synthesis companies could be required to conduct screenings in all cases.
-
Potentially harmful papers could be removed from the training data for AI systems.
-
More caution could be exercised when publishing papers with recipes for building deadly viruses.
-
These measures could help control the amount of harmful information AI systems can access and distribute.
Positive Developments in Biotech: Positive actors in the biotech world are beginning to take these threats seriously. One leading synthetic biology company, Ginkgo Bioworks, has partnered with US intelligence agencies to develop software that can detect engineered DNA on a large scale. This indicates how cutting-edge technology can be used to counter the potentially harmful effects of such technology.
Managing Risks from AI and Biotech: Both AI and biotech have the potential to be beneficial for the world. Managing the risks associated with one can also help manage risks from the other. Therefore, ensuring the difficulty in synthesizing deadly plagues protects against certain forms of AI catastrophes.
-
The important point is to stay proactive and prevent detailed instructions for bioterror from becoming accessible online.
-
Preventing the creation of biological weapons should be difficult enough to deter anyone, whether aided by AI systems like ChatGPT or not.
Source (Vox)
GPT-3 was given an IQ test and found to earn a score of 112. More recently, as reported by Scientific American, GPT-4 scored 155 on the test. This score is five points below what Einstein scored and five points above the IQ that the average Nobel laureate scores. In a few years LLMs will probably score over 200 on these tests, and once AGIs begin to create ASIs one can easily imagine them eventually scoring a thousand or more on these tests, meaning that we will probably have to devise new ones for this scale of measurement. This is just a small example of how quickly AI is developing and of how much promise it holds for our world’s future. Just imagine the kinds of problems that these ASIs will soon be able to solve that lie way, way outside of our current human ability.
Much of our advancement in the world has had to do with the application of intelligence to ethical behavior. Government, education and medicine are good examples of such advancement. Generally speaking greater intelligence translates to a better understanding of right and wrong. For decades we have had far more than enough resources to create a wonderful world for every person on the planet but we have lacked the ethical will to get this work done. The promise of AI is that very soon we will probably have more than enough ethical intelligence to finally get this done. We are welcoming a wonderfully intelligent and virtuous new world
https://www.scientificamerican.com/article/i-gave-chatgpt-an-iq-test-heres-what-i-discovered/
Artificial intelligence (AI) has made remarkable strides in recent years, particularly in the realm of computer vision. One fascinating application of AI is the generation of realistic human faces. This cutting-edge technology has the potential to revolutionize various industries, from entertainment and gaming to personalized avatars and even law enforcement. In this article, we delve into the intricacies of AI-driven face generation, exploring the methods used, the challenges faced, and the ethical considerations surrounding this emerging field.
Understanding Generative Adversarial Networks (GANs):
At the heart of AI-powered face generation lies a sophisticated technique called Generative Adversarial Networks (GANs). GANs consist of two components: a generator and a discriminator. The generator’s role is to create synthetic images, while the discriminator’s task is to distinguish between real and generated images. Through an iterative process, the generator becomes increasingly proficient at producing images that deceive the discriminator. Over time, GANs have demonstrated exceptional proficiency in generating human faces that are virtually indistinguishable from real ones.
Training Data and Network Architecture:
To create realistic human faces, AI models require a vast amount of training data. Researchers typically employ datasets containing tens of thousands of labeled images of faces. These datasets encompass diverse ethnicities, ages, and gender, enabling the AI models to capture the wide spectrum of human facial features and variations.
Deep convolutional neural networks (CNNs) serve as the backbone of AI face generation. CNNs excel at analyzing visual data by extracting intricate patterns and features. The generator network consists of multiple convolutional and deconvolutional layers that gradually refine the generated images. The discriminator network, on the other hand, uses similar CNN architecture to evaluate and classify the authenticity of the generated faces.
Progressive Growing and Style Transfer:
One notable advancement in face generation is the concept of progressive growing. Initially proposed by researchers at NVIDIA, this technique involves training GANs on low-resolution images before gradually increasing the image size. Progressive growing allows for the generation of highly detailed and realistic faces.
Another technique that enhances the quality and diversity of generated faces is style transfer. By leveraging the latent space of the trained generator network, it becomes possible to manipulate specific features of the generated faces. This allows for the synthesis of faces with desired attributes, such as adjusting age, gender, or even combining features from different individuals.
Ethical Considerations and Mitigating Bias:
While AI-generated faces hold immense potential, ethical considerations must be at the forefront of their development and deployment. One crucial concern revolves around data privacy and consent. As AI models rely on vast datasets, ensuring that individuals’ images are used with proper consent and safeguards is of utmost importance.
Moreover, there is a risk of perpetuating biases present in the training data. If the training dataset is not diverse or contains inherent biases, the generated faces may exhibit similar biases. Recognizing and mitigating these biases through careful curation of training data and algorithmic techniques is crucial to prevent discriminatory outcomes.
Applications and Future Prospects:
The applications of AI-generated human faces are vast and varied. In the entertainment industry, this technology can revolutionize character creation in movies, video games, and virtual reality experiences. It also has potential applications in facial reconstruction for historical figures and forensic facial reconstruction in criminal investigations.
Looking ahead, advancements in AI face generation could lead to breakthroughs in areas such as personalized avatars, virtual communication, and improved human-computer interactions. However, it is essential to continue research and development while maintaining ethical standards to ensure the responsible and equitable use of this technology.
Resource:
StyleGAN by Nvidia (Open Source) – GitHub » StyleGAN on GitHUB
Impressive StyleGAN Example Online » Face Generator Free
StyleGAN on Wikipedia » Wikipedia StyleGAN
Can you distinguish a real face from an artificial one? » Game with GAN Technology
Could an AI create a new religion that reinterprets current dogma and unifies humanity? Imagine an AI claiming it has established a communication link to the spiritual entity in charge of the universe, and determined that “This is what she meant to say.”
I wonder if Google is already working on that?
A few hours ago, Singapore’s global cybersecurity leader, Group-IB, identified 101,134 compromised ChatGPT accounts.
If you want the latest AI news as it drops, look here first. All of the information has been extracted here as well for your convenience.
More specifically, the credentials of over 100,000 accounts were leaked on the dark web. These compromised credentials were found within the logs of info-stealing malware sold on illicit dark web marketplaces over the past year.
The number of available logs containing compromised ChatGPT accounts reached a peak of 26,802 in May 2023, with the Asia-Pacific region experiencing the highest concentration of stolen ChatGPT credentials.
Info stealers are a type of malware that collects credentials, bank card details, and more from browsers installed on infected computers, before sending this data to the malware operator. They have emerged as a significant source of compromised personal data.
Group IB has identified the perpetrator as “Raccoon” an infamous info stealer.
What’s most interesting is that 2FA is currently paused in ChatGPT as of June 12th
So there is no way to enable extra security as of now but changing your password may be a good idea.
Full article: (link)
The idea of AI replacing hiring managers has been a topic of hot discussion. While AI can certainly play a significant role in streamlining and improving the hiring process, completely replacing hiring managers is unlikely and comes with several challenges. Here are a few points to consider:
1. Human Interaction: Hiring involves complex decision-making that goes beyond analyzing resumes and qualifications. Hiring managers often assess candidates’ soft skills, cultural fit, and potential through interviews and interactions. Human judgment and intuition are crucial in making these assessments.
2. Bias and Fairness: AI systems are only as good as the data they are trained on. If the training data is biased, the AI system may perpetuate biases in the hiring process. Hiring managers can bring awareness to bias and ensure fair evaluation of candidates.
3. Contextual Understanding: Hiring managers possess the ability to understand the specific needs and goals of the organization. They can align hiring decisions with the company’s culture, strategic direction, and long-term vision, which may be challenging for AI systems without contextual knowledge.
4. Adaptability and Flexibility: Hiring managers can adapt their approach based on the unique requirements of each role and the changing needs of the organization. They can pivot the hiring strategy, refine job descriptions, and prioritize qualities that align with evolving business objectives.
5. Candidate Experience: AI can streamline initial resume screening and automate certain aspects of the hiring process. However, the human touch and personalized communication from hiring managers contribute to a positive candidate experience, fostering engagement and a sense of connection with the company.
Read full article
I use google docs all the time for school this is super exciting
If you want the latest AI news as it drops, look here first. All of the information has been extrapolated here as well for your convenience.
Essay writing just became 100x easier. You can try the AI out with these 4 steps:
-
Navigate to Google Docs: Once you’re in Google Docs, look for the magic wand tool. (Look at the video to find the magic wand.) This is where the real magic begins. Describe the content you’re looking to generate in a few words, and Google will do the rest. The best part is that you can lengthen it, shorten it, and even change the tone to best fit your needs.
-
It’s in your hands: Now that you have your workspace set up, you can start generating any kind of content you want. It can be anything: a paper, an essay, a definition, the possibilities are endless.
-
Change Existing Text: One of the coolest features of Google Labs is its ability to edit existing text. Just select the text you’ve already written, and you can change it with one click or describe how you want to change it. For instance, you could instruct Google to “rewrite it with a formal tone.”
That’s it! Hope this was helpful.
Paper link: https://arxiv.org/abs/2306.10763
Discussion on twitter: https://twitter.com/LakshyAAAgrawal/status/1671498941009997828
Abstract: Language models of code (LMs) work well when the surrounding code in the vicinity of generation provides sufficient context. This is not true when it becomes necessary to use types or functionality defined in another module or library, especially those not seen during training. LMs suffer from limited awareness of such global context and end up hallucinating, e.g., using types defined in other files incorrectly. Recent work tries to overcome this issue by retrieving global information to augment the local context. However, this bloats the prompt or requires architecture modifications and additional training. Integrated development environments (IDEs) assist developers by bringing the global context at their fingertips using static analysis. We extend this assistance, enjoyed by developers, to the LMs. We propose a notion of monitors that use static analysis in the background to guide the decoding. Unlike a priori retrieval, static analysis is invoked iteratively during the entire decoding process, providing the most relevant suggestions on demand. We demonstrate the usefulness of our proposal by monitoring for type-consistent use of identifiers whenever an LM generates code for object dereference. To evaluate our approach, we curate PragmaticCode, a dataset of open-source projects with their development environments. On models of varying parameter scale, we show that monitor-guided decoding consistently improves the ability of an LM to not only generate identifiers that match the ground truth but also improves compilation rates and agreement with ground truth. We find that LMs with fewer parameters, when guided with our monitor, can outperform larger LMs. With monitor-guided decoding, SantaCoder-1.1B achieves better compilation rate and next-identifier match than the much larger text-davinci-003 model.
ResearchAndMarkets.com has released a new report analyzing the global AI market and making predictions for 2023.
Here are the Six Key Emerging AI Market Trends
-
Democratization of AI reduces enterprise workloads and helps jump-start machine learning projects.
-
Multimodal AI plays an important role in unlocking data’s potential.
-
Increased investment in generative AI is unleashing creative industry applications.
-
Conversational AI is emerging as a highly deployed AI technology.
-
Vendors are building edge-to-cloud integration platforms and service offerings to support data orchestration.
-
Ethical AI principles are emerging as a core aspect of implementing AI technologies.
What happened in AI this week?
1. EU approves world’s first laws regulating AI, challenges tech giants’ power
The landmark E.U. AI Act looks to protect consumers from dangerous AI applications, such as forcing tech companies to label AI-generated content.
Source: https://theconversation.com/eu-approves-draft-law-to-regulate-ai-heres-how-it-will-work-205672
2. New OpenAI updates for GPT 3.5 and 4
Updates include new function calling abilities for devs, new GPT 3.5 Turbo and GPT 4 model enhancements, and cost cuts to the current pricing. The updates aim at improving workability for developers.
Source: https://techcrunch.com/2023/06/13/openai-intros-new-generative-text-features-while-reducing-pricing/
3. UN chief backs proposal to monitor AI
During a policy implementation regarding disinformation, UN Secretary-General Antonio Guterres expressed concerns about generative AI and voiced support for a policy that creates an international AI watchdog.
Source: https://www.voanews.com/a/un-chief-considering-watchdog-agency-for-ai-/7133567.html
4. Google launches new AI-powered travel & product search feature
As part of its new Search Generative Experience (SGE), you will now get new travel and product recommendations through AI. With informative content such as “things to keep in mind when using a product.”
Source: https://techcrunch.com/2023/06/14/google-intros-new-ai-powered-travel-and-product-search-features/?guccounter=1&guce_referrer=aHR0cHM6Ly93d3cuZ29vZ2xlLmNvbS8&guce_referrer_sig=AQAAAG1Vgf9u8Yoi7OGOhvYsdhd3OUU9YvKftJRcgBg8jRTj7IctWKCEezHPa3j4Ky9nr3TzGy9Dvs6EYSRmDm6JBoh1Gp3TAaQPrLhTfBACemp2bdJUXbRbviBdubFlw0bS1hrg_1H4L5ABfGBeZQS3kqvNgXQ5WIAtDxgzFBRGUYZI
5. Google makes its Generative AI platform available to everyone
As part of Google Cloud, the company has made its Machine Learning Platform as a Service (ML PaaS) available to everyone. You can now use the Word Completion Model, Model Garden, and more.
Source: https://siliconangle.com/2023/06/07/google-cloud-makes-generative-ai-support-generally-available-vertex-ai/
6. Amazon adds new AI features to summarize product reviews
Amazon is now using generative AI to inform customers of what previous customers liked and disliked about the product, saving time in going through multiple reviews.
Source: https://www.cnbc.com/2023/06/12/amazon-is-using-generative-ai-to-summarize-product-reviews.html
7. Adobe announces AI Generative Recolor feature for Adobe Illustrator
Creatives can now use AI to change color, themes, and fonts of graphics using AI prompts. The features come in a broader push towards AI that Adobe has taken over the past few months.
Source: https://www.adobe.com/products/illustrator/generative-recolor.html
8. Adobe announces Firefly for enterprises to create custom generative AI models
Companies can now train Abode’s AI model around their branded assets, making creating designs around their brand theme and style easier.
Source: https://www.theverge.com/2023/6/8/23753564/adobe-firefly-enterprise-generative-ai-express-commercial
9. Zuck calls Meta’s new AI ‘too dangerous’ to make public
Meta has developed a highly versatile AI for speech generation called Voicebox, but will not release it to the public due to concerns about potential misuse of the technology.
Source: https://www.dailymail.co.uk/sciencetech/article-12215889/Meta-announces-new-AI-replicate-human-voices-wont-release-risks-misuse.html
10. Windows 12 will be full of AI features
We already have some AI support for the camera and microphone, but in the next version, it will be much better.
Also, Windows 12 should be able to make a lot better use of NPUs, or neural processing units, which are essentially processors that specialize in AI functionalities.
AI will be included in search, analysis, identification, and other features.
source
11. Marvel used AI to create the intro for Secret Invasion
The series is about shape-shifters that try to imitate humans, which is the exact phrase that can be used to describe AI. How convenient 🙂 You can check out the footage from the source.
- The Marvel series ‘Secret Invasion’ uses generative AI for a specific sequence in the opening credits, as confirmed by the director Ali Selim.
- Generative AI uses millions of images created by artists and photographers to train it, raising issues of using these pieces without artists’ permission or compensation, and potentially replacing actual artists with AI.
- Despite the visual appeal of AI art, there is controversy when it’s used in such high-profile projects without full understanding of its creation process and potential implications for artists.
source
12. AI can now predict pop music hits better than humans
AI can now predict pop music hits better than humans, according to researchers from the US. Scientists have utilized artificial intelligence to identify hit pop songs with an impressive 97% accuracy. This technology could render TV talent show judges obsolete and significantly reduce the costs of music production overall
Artificial intelligence (AI) programs constantly use machine learning to improve speed and efficiency. In reinforcement learning, AI is rewarded for desired actions and punished for undesired actions.
Reinforcement learning can only take place in a controlled environment. The programmer assigns positive and negative values (or “points”) to certain behaviors, and the AI can freely explore the environment to seek rewards and avoid punishments.
Ideally, the AI will delay short-term gains in favor of long-term gains, so if it chooses between earning one point in one minute or earning 10 points in two minutes, it will delay gratification and go for the higher value. At the same, it will learn to avoid punitive actions that cause it to lose points.
Examples of Reinforcement Learning
Real-world applications of AI based on reinforcement learning are somewhat limited, but the method has shown promise in laboratory experiments.
For example, reinforcement learning has trained AI to play video games. The AI learns how to achieve the game’s goals through trial and error. For example, in a game like Super Mario Bros., the AI will determine the best way to reach the end of each level while avoiding enemies and obstacles. Dozens of AI programs have successfully beaten specific games, and the MuZero program has even mastered video games that it wasn’t originally designed to play.
Reinforcement learning has been used to train enterprise resource management (ERM) software to allocate business resources to achieve the best long-term outcomes. Reinforcement learning algorithms have even been used to train robots to walk and perform other physical tasks. Reinforcement learning has also shown promise in statistics, simulation, engineering, manufacturing, and medical research.
Limitations of Reinforcement Learning
The major limitation of reinforcement learning algorithms is their reliance on a closed environment. For example, a robot could use reinforcement learning to navigate a room where everything is stationary. However, reinforcement learning wouldn’t help navigate a hallway full of moving people because the environment is constantly changing. The robot would just aimlessly bump into things without developing a clear picture of its surroundings.
Since this learning relies on trial and error, it can consume more time and resources. On the plus side, reinforcement learning doesn’t require much human supervision.
Due to its limitations, reinforcement learning is often combined with other types of machine learning. Self-driving vehicles, for example, use reinforcement learning algorithms in conjunction with other machine learning techniques, such as supervised learning, to navigate the roads without crashing.
President Biden emphasizes the importance of ensuring safety in AI before its deployment, urging for bipartisan privacy legislation and new safeguards for this emerging technology. He calls for stricter limits on personal data collection, bans on targeted advertising to children, and the requirement for companies to prioritize health and safety.
Biden’s Stance on AI Safety and Privacy: Biden advocates for pre-release safety assessments of AI systems. He stresses the risks that unsafeguarded technology can pose to society, economy, and national security.
Effect on Social Media and Advertising: Biden identifies potential harm from powerful technologies like social media, especially without adequate safeguards.
-
He notes the need for strict restrictions on personal data collection.
-
He advocates banning targeted advertising to children.
-
He insists on companies prioritizing health and safety.
Discussion with Tech Leaders: Biden met with prominent figures in the AI and education sectors, including leaders from Khan Academy, the Center for Humane Technology, and the Algorithmic Justice League among others.
Efforts Towards Privacy and Security Protections: White House Chief of Staff Jeff Zients oversees the development of additional steps the administration can take on AI.
-
Zients notes the cooperation of AI companies in introducing privacy and security commitments.
-
Vice President Kamala Harris plans to convene civil rights and consumer protection groups for AI discussions.
Involvement of Major AI Firms: The administration seeks to involve leading AI companies in its efforts.
-
Meetings have been held with CEOs of major firms like OpenAI, Microsoft, and Alphabet.
-
These companies have agreed to participate in the first independent public evaluation of their systems.
Prospective Regulatory Measures: The administration looks towards broader regulatory initiatives for AI, involving multiple federal agencies.
-
The Commerce Department considers rules for mandatory AI model certification before release.
-
The Federal Trade Commission monitors AI tool usage.
-
Congress scrutinizes AI technology, with Senate Majority Leader Chuck Schumer set to outline his vision for AI’s potential and its safeguards.
Source (Bloomberg)
Just recently, a paper went viral on Twitter which suggested GPT-4 scored 100% on the MIT EECS + Math curriculum (link). However, the results showcased in the paper proved “too good to be true” and some post analysis reveals major issues with different aspects of the study.
Dataset Issues
The authors state that GPT-4 was able to score 100% on a randomly selected set of 288 questions. However, on close inspection of the data-set, it was found that the data-set contained a number of questions (roughly 4%) that were “unsolvable”, such as:
Eg: “At the command prompt, type: traceroute 18.31.0.200 Describe what is strange about the observed output, and why traceroute gives you such an output. Refer to the traceroute man page for useful hints. Copy/paste any of the relevant portions of output below.”
The true answer can not be found given this information, because the context is too limited, and without access to an interactive terminal (no such access was given in this work), it would be impossible for an LLM agent to answer.
Information Leak in Few Shot Examples
There was discovered evidence of significant data leakage within the few shot examples provided for the model. Many were nearly identical to the problems themselves, essentially giving the model the answers.
Grading Methodology
There were problems with the paper’s grading methodology as well. The system checked with GPT-4 using the original question, ground solution, and the model’s own answer. This has the potential for the AI to produce inaccurately high self-assessment scores, especially in technical fields, where it may have hidden misunderstandings.
Second, there are risks of data leakage in the prompt cascade approach used in the paper. The approach provides binary feedback based on the ground truth, and the system reprompts until the correct answer is reached. This issue is particularly significant in multiple-choice problems (representing 16% of the test set), where unlimited attempts almost guarantee the right answer, comparable to a student receiving continuous feedback about the accuracy of their answers until they get them right.
There was an extensive analysis done by 3 MIT EECS seniors which can be found here: (link)
There are a good amount of quality AI chat alternatives out there besides ChatGPT and some even offer GPT-4 for free! Here’s a list of alternative chatbots to try out (I’ve tried all of these not some bs list):
Perplexity: “The first conversational search engine” (GPT-3.5 Free / GPT-4 Paid $20 a month)
Bing: Microsoft’s Chatbot with multimodal capabilities. (GPT-4 Free)
Poe: Quora’s AI app with multiple models (GPT-3.5 Free / GPT-4 free with ‘limited access’)
AgentGPT: “Autonomous AI agent” Give one prompt and it will run continuously until finished. (GPT 3.5 Free / GPT-4 API access required) sign up for GPT-4 API waitlist here
HuggingFace: Largest open source AI community find thousands of different open source projects (Free site)
Ora: Access community LLM’s or build your own (GPT-3.5 Free / GPT-4 Free) Direct link to free GPT-4
Inflection Pi: A personal AI chatbot (not meant for research purposes) (Free site) … unsure what model I have seen conflicting information I believe it’s GPT-3.5
Nat.dev: Use GPT-4 in playground and compare to other models (GPT-4 $5 credit fee)
Merlin: Access GPT-4 chatbot in any browser (GPT-4 limited free plan / GPT-4 unlimited starting at $19 a month)
These are all credible chatbots that have been running for months the majority do require email signups however. Hope this helps!
The Telegraph reports
Victims should be able to use artificial intelligence (AI) to find out their chances of success in court claims, Britain’s top judge said.
Lord Burnett of Maldon, the Lord Chief Justice, cited AI technology already being used in Singapore that allows road traffic accident victims to tap in the details and find out within 10 minutes the probable outcome of any litigation they might decide to take.
The system helps victims decide whether it is worth pursuing in the courts based on the AI technology’s analysis of the current law and case precedents. This can form the basis for a swifter settlement without a victim resorting to legal proceedings.
Lord Justice Burnett told peers on the Lords constitution committee: “It is not binding, you can issue proceedings, but it is the sort of thing that would be of some use. So I think AI is something which we want to be looking at to enhance access to justice.
“The administration of justice and the courts should try to harness developments in technology which enhance the rule of law and what we do. We should never be the slave to it, but undoubtedly there will be ways in which artificial intelligence can be used to increase access to justice, for example.”
Read more: https://www.telegraph.co.uk/news/2023/06/20/ai-victims-court-case-win-lord-chief-justice-britain/
AI Daily News on June 21st 2023
Google’s First Two AI-powered Ad formats
With this new Google Ads Update. Google is bringing faster Ad set creation for demand generation Ads.
And new updates to Youtube Ad campaign creation where these Demand Gen Video Ads with AI-powered lookalike audiences are performinv great with Beta testers like Arcane & Samsung.
Tiktok’s AI Ad Script Generator
Tiktok product marketing team announced a new Advertising feature for marketers this week. You can visit the video tutorial to see this new AI Ad tool in the action. It is also available in Tiktok Ads manager now to all advertisers.
Supermetrics launched AI integration with Google Sheets
The platform, recommended by Google workspaces for marketing data. Launched new GPT integrations with AI & GPT 4 for their Google Sheets Integration.
Meta & Microsoft Sign a pact to responsibly use AI
Partnership on AI association shared that both companies have signed a pact with them about AI usage. Both of these platforms will be following framework introduced by PAI’s framework to partner for non-profite AI research & projects.
Ogilvy is asking other agencies to label AI-generated content
As AI-influencers are taking over, Oglivy is requesting agencies & policymakers to enforce brands toAI-generated influencer content. As they believe influencers are trusted figures in marketing and not labeling AI-influencers breaks consumer trust.
Microsoft AI Ads
During a latest event about AI advertising, Microsoft’s VP of Ads shared her insights on where microsoft is heading with AI Ads for Bing Chat & Search.
For Context, so far Microsoft have introduced around 5-8 new AI related product updates. The Ad platform is changing fast and they are embracing AI Ads faster than Google & Amazon.
Adobe’s new AI Update for Illustrator
Adobe firefly is already making a huge spark with new features. And last week, the platform launched new graphic design Generative recolor feature to Adobe illustrator.
This new feature is a great addition for brand designers & marketers looking to build new brand identity.
Bing Chat tests Visual Search
Bing’s next step is to takeover the Google Lens product. Microsoft is testing Visual search & photo recognition feature for Bing Chat. Here are the first looks
This feature will have a major impact on Google & Pinterest’s visual search features.
- Meta has developed a new AI system, “Voicebox”, which can generate convincing speech in a variety of styles and languages, and perform tasks such as noise removal, outperforming previous models in terms of speed and error rates.
- Despite potential benefits such as giving a voice to those who can’t speak, enabling voice inclusion in games, and facilitating language translation, Meta has decided not to release the model due to concerns over misuse and potential harm, like unauthorized voice duplication and the creation of misleading media content.
- To manage risks, Meta has developed a separate system that can effectively distinguish between authentic speech and audio generated with Voicebox, but it remains cautious about releasing Voicebox to the public, emphasizing the importance of balancing openness with responsibility.
-
Mark Zuckerberg shared that they have built one of the best AI speech generation product. But the product is too dangerous for public access, it will not become available anytime soon. May be in next few years as revealed by Mark zuckerberg.
You can check the Meta’s Voicebox release statement here.
Pixar uses Disney AI
Revealed by a recent wired article. Pixar utilised Disney’s AI technology for their upcoming Elemental Movie.
OpenAI plans app store for software
OpenAI is planning to launch a marketplace where developers can sell their AI models built on top of ChatGPT, according to sources. The marketplace would offer tailored AI models for specific uses, potentially competing with app stores from companies like Salesforce and Microsoft, while expanding OpenAI’s customer base. This platform could serve as a safeguard against reliance on a single dominant AI model. And It’s not clear whether OpenAI would charge commissions on those sales or otherwise look to generate revenue from the marketplace.
If OpenAI proceeds with this, it could herald a new era in the AI industry. It would provide a platform for businesses to not only create but also monetize their AI models, fostering a more collaborative and innovative environment.
While the idea is promising, it’s not without potential hurdles. Questions around intellectual property rights, quality control, and security could arise. How will OpenAI ensure the quality and safety of the models being sold?
This marketplace could potentially accelerate the adoption of AI across various industries. By providing a platform where businesses can purchase ready-made, customized AI models, the barrier to entry for using AI could be significantly lowered.
Elon Musk repeats call for halt on AI development
Elon Musk reiterated his belief that there should be a pause in the development of AI and called for regulations in the industry. He expressed concerns about the potential risks of digital superintelligence and emphasized the need for AI regulation.
Xi Jinping says China welcomes US AI tech
Chinese President Xi Jinping held discussions with Bill Gates regarding the global growth of AI and expressed his support for U.S. companies, including Microsoft, bringing their AI technology to China.
EU lawmakers vote for tougher AI rules as draft moves to final stages
European Union lawmakers have agreed on amendments to draft AI rules that would ban the use of AI in biometric surveillance and require disclosure of AI-generated content by generative AI systems like ChatGPT. The proposed changes could lead to a clash with EU countries opposing a complete ban on AI in biometric surveillance, and the amendments also include requirements for copyright disclosure, distinguishing deep-fake images, and safeguarding against illegal content.
vLLM: Cheap, 24x faster LLM serving than HF Transformers
The performance of LLM serving is bottlenecked by memory. vLLM addresses this with PagedAttention, a novel attention algorithm that brings the classic idea of OS’s virtual memory and paging to LLM serving.
It makes vLLM a high-throughput and memory-efficient inference and serving engine for LLMs. vLLM outperforms HuggingFace Transformers by up to 24x (without requiring any model architecture changes) and Text Generation Inference (TGI) by up to 3.5x, in terms of throughput.
Google DeepMind’s RoboCat pushes the boundaries of robotic capabilities
Google DeepMind has created RoboCat, an AI model that can control and operate multiple robots. It can learn to do new tasks on various robotic arms with just 100 demonstrations and improves skills from self-generated training data.
RoboCat learns more quickly than other advanced models because it uses a wide range of datasets. This is a significant development for robotics research as it reduces the reliance on human supervision during training.
Cisco introduces networking chips for AI supercomputers that would compete with offerings from Broadcom and Marvell Technology.
They revealed that chips are from its SiliconOne series, and currently being tested by five out of the six major cloud providers. Notable cloud players like AWS, Microsoft Azure, and Google Cloud, which together dominate the market for cloud computing.
Teleperformance signed a multi-year $185M Azure Cloud commitment with Microsoft to launch GenAI platform.
Through the collaboration, the objective is to provide Microsoft Cloud infrastructure solutions to clients. Teleperformance will also use Microsoft Azure AI, to launch TP GenAI, a new suite of AI solutions for faster and improved business processes.
- OpenAI has lobbied the EU to soften proposed AI regulations, arguing that general-purpose AI systems (GPAIs) like ChatGPT shouldn’t be considered “high risk” under the forthcoming EU AI Act, which would impose strict safety and transparency requirements.
- Despite initial secrecy, OpenAI supported the inclusion of “foundation models” (powerful AI systems used for various tasks) in the AI Act, which demands more transparency, including disclosing whether copyrighted material has been used for training AI models.
- The EU AI Act, with some of OpenAI’s proposed changes incorporated, has been approved by the European Parliament, but still needs to go through a final “trilogue” stage before it comes into effect, a process expected to take about two years.