AI: Weekly Summary. RTZ #374

1. OpenAI and Apple deals in the offing: Reports keep coming on OpenAI and Apple announcing a deal around GPT powering Siri at Apple’s WWDC conference on June 10. Bloomberg has more details on possible details and features, including advanced AI multimodal capabilities for Siri. Apparently Apple remains in discussions with Google as well on Gemini powering iPhone AI features going forward. Important detail given Google already paying Apple $20 billion plus a year for default Search positioning in 2.2 billion plus Apple phones, macs, and devices. Apple also needs to balance introducing AI features with its unique commitment to user privacy and trust. 

2. Elon Musk funds xAI with investor round: Elon Musk completed its $6 billion fund raise for xAI at a $24 billion post valuation, and advanced plans for an AI ‘Dojo’ Compute Gigafactory, powered by Nvidia, Oracle, and others. xAI plans to leverage the range of data sources across Elon Musk’s companies, from Tesla to X/Twitter. xAi of course is the latest of big tech companies expanding AI datacenter infrastructure to the benefit of Nvidia, TSMC, and others, including Microsoft/OpenAI, Google, Amazon, Meta and others. Apple has yet to officially throw its hat into the AI datacenter ring at scale. More on xAI efforts here.

3. Google’s AI errors in context: Google responded to the recent controversy over its Gemini Ai driven Search results. The company explained ‘AI Overviews’ virtual mistakes and updates’, and defended its accuracy. The response reassures that Google’s ‘AI Overviews’ don’t ‘hallucinate’ or make things up ‘like other LLM products’. But underlined that when the results are incorrect, they reflect other factors at work. Whole piece  by the head of Google Search Liz Reid, worth a read given the importance of this issue for LLM AIs going forward. More here.

4. Meta builds on TikTok type algorithm:  Meta continues to redo its ‘For You’ Feed algorithms, to feature content beyond the user’s social graph, and more like Bytedance TikTok’s approach.  The focus is more on user interests rather than user networks. Goal of course, is to appeal more to younger demographics across Instagram, WhatsApp, and Facebook. Meanwhile, TikTok is cloning an US algorithm to separate the AI technologies used at parent Bytedance. All ahead of course of ongoing US regulatory and legislative efforts on banning and/or encouraging a TikTok divestiture. Separately, Meta is also continuing to use Instagram, Facebook and other data to train its AI models, with an opt-out in Europe.

5. Enterprise cloud growth stumbles: The Enterprise cloud software sector saw some weakness this week, led by results at Salesforce, Dell, Workday, MongoDB and others. Many dropping 10-20% or more. Big question is if enterprises are holding off on big software commitments given ongoing AI experiments and deployments. It’s important to note that the above corrections are after strong performance in these stocks all year to date. Dell alone was up over 100% this year, before this week’s results. More on the issues here.

Other AI Readings for weekend:

1. OpenAI is apparently rebuilding its  robotics team, after disbanding it in 2015. Foundation robotics is a key next focus for LLM AI innovation in the physical world, especially by AI infrastructure leading Nvidia with its GR00T, Isaac, and other AI robotic efforts.

1. LLM AIs like GPT-4 do financial analysis better, according to new research, highlighting inroads into white collar work vs blue collar. More here.

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)