AI: Weekly Summary. RTZ #353

1. OpenAI Voice: Despite expectations that OpenAI would release it’s long awaited AI Search to compete with Google I/O Developer Conference on Tuesday, 5/14, it looks like the plan is to roll out their also long awaited AI Voice product, and “some ChatGPT and GPT-4 updates”. This has been on a slow roll and long testing road-map due to long standing concerns of its misuse, especially in an Election year. Also possible are incremental upgrades to GPT 4, before releasing GPT-5 either later this year, or early next. More on other OpenAI product roadmap possibilities here.

   

2. Microsoft MAI-1 LLM as alternative to OpenAI GPT-4: Microsoft’s new AI division under former Inflection founder Mustafa Suleyman is working on a large 500 billion parameter LLM named MAI-1. Speculation is rife on Microsoft’s motivations, but at the very least it leverages Microsoft engineers and compute resources with other LLM models beyond its core partnership with OpenAI. More here.

   

3. Google AlphaFold 3 and I/O Developments: Google DeepMind introduced AlphaFold 3, to better predict protein structure, DNA, RNA and other elements. This is a meaningful scientific advancement, and contributes meaningful to AI tools in healthcare. Expect other major AI announcements of course at Google’s annual I/O Developer conference this Tuesday 5/14. Google also makes headway reducing AI operational costs with their formidable TPU training and inference infrastructure, as discussed here.

   

4. Apple ACDC AI Inference chip: Apple is apparently readying its own AI server chips dubbed ACDC (Apple Chips for Data Center), designed on Apple SIlicon, and made by TSMC. The focus seems to be on Inference loads, as opposed to train their own models for now. Of course there are separate reports that the company is also potentially talking to OpenAI and/or Google to use their ChatGPT/GPT and/or Google Gemini models for Siri and other Apple services. The timing would be to have these announcements ready for Apple’s WWDC Developer conference on June 10. More here.

   

5. Hollywood Concerns over AI Use: There is an increasing debate over the pros and cons of AI tools being used in Hollywood, with good arguments on both sides. Particular focus of course is on AI use disclosure and labeling. Thus all of course is in addition to the ongoing debates over compensation to artists around AI as well. More here. 

   

Other AI Readings for weekend:

1. Continued angst over Evaluating LLM AI models, as they continue to improve and best each others’ capabilities at a rapid clip. The challenges are likely to get harder in the near term.

   

2. Took a crack at discussing AI Plurality vs AI Singularity, especially as AI machine to machine (m2m) traffic is likely to increase exponentially with Smart Agents and Agentic Workflows.

   

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)