Tuesday, December 03, 2024

There is more in heavan and earth than are dreamed of in your philosophies.

 The big AI putsch now is LLMs. Everybody having the server sized chops has one. Hugging Face is a popular site.

Beyond LLMs, everybody who is anybody is designing an AI chip. 

search = ai chips being designed  

Returned items include one from Google Deepmind, and this item. As examples. Have a look.

Now, chip design is one thing, fabrication another, and what of "foundries."

TSMC and Intel are building foundry capacity stateside, in part subsidized by the federal government. Those will be leading fab outlets. TSMC being a Taiwan firm, is a part of China, but with wrinkles. See, e.g., here and here. And this is Biden lame duck China restraining, before Trump takes on his plans.

Those with chip designs will have to queue up, and FIFO will rule unless the foundries want litigation. 

And Elon will be well positioned, aside from others. If they bottleneck things with more demand than foundry capacity, it may get political. There, Musk has become close with the Trump administration, having DOGE duties, and he will be accorded industry fairness, if not favoritism, from it. So - who gets shrunk out of jobs in the administrative state, while Elon has multiple pecuniary interests? 

Interesting question? Ya betcha.

We live in interesting times. Where it all goes is guesswork, but politics suggest Elon will have his share of whatever he wants to bite off and chew, his positioning making that one hell of a good guess you could gamble on.

___________UPDATE__________

Amazon, yesterday's news, joins the LLM frenzy -

The Seattle-based company is stringing together hundreds of thousands of its Trainium2 semiconductors into clusters that will make it easier for partner Anthropic to train the large language models required for generative AI and other machine learning tasks.

Their own chip design, using Anthropic's LLM technology; and 

AWS, the largest seller of rented computing power, runs many of the servers that other companies rent to train artificial intelligence applications. AWS also makes models built by other companies, including Anthropic’s Claude and Meta Platforms’s Llama, available to AWS customers. But the company has yet to produce a large language model widely seen as competitive with OpenAI’s most advanced GPT models. 

Prior Amazon-built models released in the last two years, called Titan, were generally smaller in scope. The Nova models, some available now and others next year, include a “multimodal to multimodal” version that can take text, speech, images and video as inputs and generate responses in each mode.

Amazon, Jassy said, would continue to both develop its own models and offer those built by others. “We are going to give you the broadest and best functionality you can find anywhere,” he said.

As noted online, Musk is suing OpenAI for attempted monopolization of the market niche. Microsoft a co-defendant. BBC story, and pdf copy, 107p complaint. Also, TechCrunch.

Bezos, on his super yacht, might be holding Elon's coat on that effort.