AI: Weekly Summary. RTZ #332

AI: Weekly Summary

…week ending April 19, 2024

1. Meta Shifts Gears with Meta AI & Llama 3: Meta founder and CEO Mark Zuckerberg announced a big shift into ‘smart agent’ and ‘chatbot’ oriented ‘Meta AI’ service, now powered with its latest version of Llama 3. The company also underlined its continued support for open source LLM AI models, highlighting that Llama 3, when available later this year in its 405 billion parameter size later this year, will be the largest open source LLM AI out there. In the meantime, an 8 billion and 70 billion sized Llama 3 models are available today, supported by both Google and Bing Search for accurate, real-time information. The new models seem to rank favorably with competing offerings. A more detailed dive here.

2. Microsoft invests in Mideast AI company G42: Microsoft signed an investment to partner and invest $1.5 billion in UAE AI company G42, a deal supported by the US. Microsoft President Brad Smith joins G42’s board, while G42 gets to put the UAE’s AI models on Microsoft Azure cloud services. The US continues to encourage Mideast countries and its tech investments to shift away from China, and this deal is likely the first in others to follow. More detail here.

3. Google reorganizes AI across Google devices: Google CEO Sundar Pichai moved to further consolidate its Deepmind and Research teams around AI, this time relocating its AI Responsible AI teams from Research to Deepmind. Also, a ‘new unified Platform and Devices team, will bring together efforts across hardware, software and AI teams at Google’, including the company’s Android, Chrome, search and photos businesses.  The stated goal is to “help the company develop AI products and services faster and more efficiently. The moves continue to highlight the company’s efforts to streamline its ‘wood behind the Arrow’ with Gemini driven AI, a distinct contrast to historical approach of having multiple efforts flourish concurrently at the company. 

4. Alternative LLM AI companies emerging: While the big LLM AI companies continue to furiously innovate with next gen models, alternative AI strategies are being rolled out by startups. Companies like Symbolica, Physical Intelligence and others follow in the steps of Sakana AI last year, to try different approaches to Generative AI, in many cases blending pure Gen AI with traditional logic based software systems. The ‘zig-zag’ strategies try to focus on alternative approaches to Scaling AI, besides pouring ever larger amounts of compute, power and capital, which is the current ‘Hyperscale’ approach. Additional details and context here.

5. Tit for Tat on TikTok: The US House moves towards accelerating the TikTok ban/divestiture, bundled in a bill coupled with the Aid to Ukraine and Israel. Likely in response, China responds by making Apple remove Meta’s Threads and WhatsApp apps in China. Continues the tech and geopolitical ‘threading the needle’ tit for tat game between the US and China.

Other AI Readings for weekend:

1. Stanford University’s Human Centered AI group published its 7th Annual AI Index Report for 2024, with 9 key takeaways. Most are moving in the positive direction.

   

2. France’s Mistral gets ready to raise additional capital at a $5 billion valuation, accelerating European AI aspirations.

Thanks for joining this Saturday with your beverage of choice. 

Up next, the Sunday ‘The Bigger Picture’ tomorrow. Stay tuned. 

(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here)