AI: Weekly Summary. RTZ #339

1. Tech Quarterly Season Starts with Meta, Microsoft, Google et al: Earnings season is back, with three of the ‘Magnificent 7’ Big Tech out with solid AI-driven results. While Meta stock retreated a bit post results (up almost 40% this year), Microsoft and Google were up. Google enthusiasm also reflects a first-ever dividend, so there’s that. Overall, the ‘AI Hyperscalers’ are generally boosting their AI capex by tens of billions each going into the second half of 2024 and next year. And the street is generally supportive. Amazon, Oracle, and others up next week. More here. 

2. Positive Reception to Meta Llama 3: Last week’s release of Meta and CEO Mark Zuckerberg’s next generation open source LLM AI Llama 3 (in 3 size flavors) got positive reviews from tech trades and developers this week. Their release of Meta AI, a multimodal ‘ChatGPT’ service also saw mostly positive reviews, with expectations of fixing and delivering a range of items. The trend continues a vigorous slate of LLM AI releases this year from a range of other companies (see below).

3. LLM AI Models catch up to OpenAI GPT-4, but GPT-5 looms: The week also saw a robust slate of both open and proprietary LLM AI update and upgrade releases, that all compare favorably with the market leader OpenAI GPT-4. Snowflake announced their open source Arctic LLM, following competitor Databricks’ release of their DBRX LLM AI a few days ago, Microsoft releases Phi-3 ‘Small Language Model (SLM), joining the lineup with Google’s Gemma, Anthropic’s Claude 3 Haiku, and others in the growing SLM LLM AI category. All of them neck and neck on LLM AI eval scores, at least until OpenAI’s GPT-5 comes on the scene possibly this year or early next.

3. US Chip Fab Momentum continues: With ever more powerful LLM AI models, come ever more powerful, and just more AI GPU chips, Systems, and Infrastructure. On that front, the momentum continues as the US Chip Act gets deployed with billions going to Chip  and Chip Foundry companies foreign and domestic, to make investments in the US in the tens of billions, spanning years to build out the facilities. Representative tale of the deep technical and business model involved, is around Intel, as it unrolls its Intel Foundry strategy. Global Foundry leader TMSC with over 60% share is also of course building fabs in the US, along the current number two Samsung, with a 13% global share. HBM AI memory systems providers like SK Hynix from South Korea are also investing billions in the US on chip facilities. More here.

5. New focus on ‘Agentic AI Workflows’: Besides releasing LLM AI models that go toe to toe with OpenAI’s GPT-4, the other focus by AI companies large and small is in the ‘Agentic AI Workflow’ space. This is complementary to the industry’s efforts around AI ‘Smart Agents’ and Companions, and targeted at both end-user and business markets. Google, Microsoft, OpenAI and most other tech companies are aggressively focused in this area, with a lot of AI technical wood to chop. More here.

Other AI Readings for weekend:

1. New US AI Safety Board at Department of Homeland Security, with OpenAI, Microsoft, Google and others participating. Focus is on ‘deploying AI safely within US critical infrastructure.

2. In the ‘more the merrier’ department, Elon Musk’s xAI finalizing $6 billion raise on $18 billion pre-money for Grok and more.

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)