Good morning.

Welcome to the Blacklynx Brief.

As you will read below - AI is now able to beat any human on tests like the SATs or the bar exam. It also outpaces humans in image classification, pattern recognition, and reading comprehension.

It already had our number in complex games like Chess, Go, and even Starcraft 2.

This is the new reality.

We’re outsmarted.

They say knowledge is power, and these AI engines have the potential for true power.

And while I understand why that could scare you, there is also huge promise for us humans.

AI could help us cure cancer, solve climate change, and get us to other planets if it keeps getting smarter.

It could also wipe us off the face of the planet if this power is misdirected.

If you think this is science fiction, Elon Musk and others have voiced similar concerns. Say what you want about Elon Musk—a moron he is not.

Musk and others, like Lee Cronin, have said AI is more dangerous than nuclear bombs. Those people are visionaries, we should listen to them.

I understand the people who think this is all overblown. Or think that AI is smoke and mirrors. LLMs, in a sense, ARE a party trick. But if you ‘re a trusted subscriber, you can see the evolution in last few weeks and months.

So..

Welcome to the Blacklynx Brief

AI NEWS
Tesla , the AI company

• Stanford University's 2023 AI Index report shows that artificial intelligence now outperforms humans in most major benchmarks, such as image classification and reading comprehension. The rapid progress of AI has rendered many existing benchmarks obsolete, prompting the development of new tests to gauge AI capabilities. While closed-source models currently lead the industry, which is dominated by a few major players, LLMs are improving in accuracy and reducing errors like "hallucinations." This report underscores the swift and significant advances in AI, setting the stage for even more dramatic developments in the near future.

  

• Meta has launched Llama 3, an advanced open-source language model, featuring versions with 8 billion and 70 billion parameters that outperform competitors in various benchmarks. Llama 3, which utilized a dataset seven times larger than its predecessor, will soon be expanded with a 400 billion parameter version, rivaling GPT-4. This model is integrated into Meta's platforms like Facebook and Instagram, making cutting-edge AI accessible to its billions of users.

• Microsoft has introduced VASA-1, an AI model that creates highly realistic talking head videos from just one photo and an audio clip. This technology not only syncs lip movements but also animates nuanced expressions and even singing performances. VASA-1 comes with customizable features, enhancing its potential for use in virtual avatars, gaming, and animation. However, its ability to produce lifelike deepfakes also raises important ethical and security concerns.

• A recent University of Cambridge study has shown that OpenAI's GPT-4 can analyze eye conditions and suggest treatments nearly as well as expert ophthalmologists. GPT-4 scored better than trainees and junior doctors in a test involving multiple-choice questions. Despite being conducted in mid-2023, the results suggest that current language models may have even more advanced capabilities. This finding highlights the potential for AI to enhance healthcare accessibility, particularly where specialist care is scarce.

• Hollywood talent agency CAA has launched an initiative called CAA Vault, which involves creating AI clones of their A-list clients. These digital replicas can be used for reshoots, dubbing, and as stunt doubles, enhancing creative opportunities and potentially reducing production costs. CAA aims to extend this technology across the industry, signaling a significant shift as AI capabilities improve. This development could fundamentally alter how productions are made and how actors are utilized in the entertainment industry.

• Adobe has introduced a major update to its Firefly image generation model, enhancing image quality, speed, and the range of styles it can produce. This upgrade also includes new tools for more precise control in image creation, along with significant improvements to Photoshop's AI features. These advancements are part of Adobe's effort to integrate powerful AI into its leading software, positioning Firefly as a key player in the image generation field and transforming creative processes.

• Apple has acquired Datakalab, a Paris-based AI startup known for its innovative data compression and image analysis technologies. This move, part of Apple's broader strategy, enhances its on-device AI capabilities, setting the stage for significant announcements expected at the upcoming WWDC event. The acquisition underscores Apple's commitment to advancing AI within its devices, potentially transforming user experiences with new AI features in the next iPhone release.

• Microsoft has introduced a new series of compact AI models called Phi-3, which includes three versions with varying sizes but exceptional performance capabilities. The smallest, Phi-3-mini, holds 3.8 billion parameters yet performs comparably to larger models like Mixtral and GPT 3.5, and can even run directly on mobile devices

• Profluent has launched OpenCRISPR-1, the first open-source AI-developed gene editor, designed to edit the human genome more efficiently. By training AI on a vast array of CRISPR data, Profluent created a tool that performs as well or better than natural alternatives. Open-sourcing OpenCRISPR-1 aims to democratize access to advanced gene-editing technology.

• During Tesla's 2024 Q1 earnings call, Elon Musk emphasized that Tesla should be viewed primarily as an AI and robotics company, rather than just an automaker. He highlighted the Optimus robot, claiming it will be the most valuable aspect of the company, outstripping all other projects. Musk also shared that Optimus is expected to perform tasks in Tesla factories by year-end and be available for external sale by the end of 2025. Despite Tesla missing earnings expectations, the stock price increased after these announcements.

• Biotech firm Moderna is expanding its partnership with OpenAI to incorporate AI across its organization, aiming to speed up the development of mRNA treatments. Moderna's usage of custom AI models includes a variety of tasks, such as optimizing drug dosages and improving branding and communication efforts. The company has deployed 750 specific AI models and engages in an average of 120 AI-driven conversations per employee each week. CEO Stephane Bancel expressed ambitions to use AI to enhance every business process, effectively allowing their 3,000-person team to operate as if it were 100,000 strong.

Quickfire News

• Apple is developing an on-device LLM for improved privacy and faster AI responses on next-generation devices.

• Meta's Llama 3 has surpassed GPT-4 in English prompts on the LMSYS LLM leaderboard, marking a significant achievement for open-source LLMs.

• Hugging Face has launched Open Medical-LLM, a benchmark for AI model evaluation in healthcare.

• TED Talks has shared a promotional video for TED2024, created by OpenAI's unreleased Sora model.

• The Canadian government is allocating $50 million to retrain workers affected by AI, as part of a larger $2.3 billion initiative to boost AI adoption.

• TikTok is working on a new AI-based text-to-speech feature that allows users to record and clone their voices in 10 seconds for video use.

• Microsoft has hired Jason Taylor, previously with Meta, as deputy CTO to push forward the development of new AI technologies.

• Cleveland Clinic researchers have utilized AI to investigate the relationship between gut bacteria and Alzheimer's disease, aiming to predict protein interactions that may influence the disease.

• United Airlines' CIO, Jason Birnbaum, reported that the airline is incorporating AI throughout its operations, including in customer service via chatbots and automated announcements for pilots.

• Xaira Therapeutics, which includes former Stanford President Marc Tessier-Lavigne as CEO, has launched with $1B in funding. The company aims to use AI to revolutionize drug discovery.

• Nvidia is acquiring Run:ai, an Israeli software company that specializes in AI computing resource management, in a deal valued at $700M.

• Meta's Q1 earnings report shows significant profit and revenue increases, attributed to major expansions in AI investments.

• AI startup Cognition announced a new funding round that values the company at over $2B. This valuation comes despite Cognition having no revenue and mixed sentiment surrounding its AI coding agent, 'Devin'.

• Augment, backed by former Google CEO Eric Schmidt, emerged from stealth with $252M in funding. It has launched an AI-powered coding platform that competes with GitHub Copilot.

• Anthropic has released new AI alignment research. The company has developed a method to help detect when AI systems might act against instructions, enhancing safety measures.

• A coffee roastery in Finland has introduced 'AI-conic', a new blend developed with the help of an AI model that analyzed desirable flavor profiles.

• Concerns have been raised about the spread of viral deepfake videos featuring Bollywood actors criticizing Indian PM Narendra Modi, which could influence the upcoming election by spreading misinformation.

• Perplexity AI has secured $63 million in funding, valuing the company at over $1 billion for its AI search engine technology.

• Walmart is implementing 19 autonomous FoxBot electric forklifts in four distribution centers to improve efficiency following a 16-month successful pilot.

• Llama 3's capabilities have been enhanced through open-source training, with a new version by Matt Shumer doubling the context window of its standard release.

• Elon Musk announced on X that the integration of xAI’s Grok chatbot into Tesla vehicles is in progress.

• Microsoft executive Mustafa Suleyman, during a recent TED Talk, described AI as a 'new kind of digital species' and predicted that the next decade will be the most productive in human history.

Closing Thoughts

That’s it for us this week.

As you can see above - we had our first sponsor this week. And to celebrate this there was a new section with quickfire news.

Just like the field of AI , we are expanding rapidly.

Have a nice weekend.

If you want a version of this newsletter with pictures etc. Be sure to subscribe below:

https://www.blacklynxbrief.com/subscribe