- IBM’s CEO walked through some napkin math on data centers— and said that there’s “no way” to turn a profit at current costs.
- “$8 trillion of CapEx means you need roughly $800 billion of profit just to pay for the interest,” Arvind Krishna told “Decoder.”
- Krishna was skeptical of that current tech would reach AGI, putting the likelihood between 0-1%.
IBM has a surprisingly sane approach to LLMs:
Small models
Economically trained
Apache licensed open weights
Geared for tool usage/RAG
Well documented
Legal, licensed training data
Open experiment artifacts, including A/B tests
See: https://huggingface.co/ibm-granite
No nebulous promises, no existential hype, no scorching the environment, no underbaked user facing disasters. Just plain locally runnable tools.
It’s so sensible it hurts. This is what all “AI” companies should be doing, albeit with a little more budget, and more modern architectures than dense GQA.
“Never interfere with an enemy while they are making a mistake”
We can quibble about if they are true enemies but these AI companies are certainly not our allies.
I hope to live long enough to see Facebook/Google/Amazon be acquired by an investment firm and sold for parts.
Meh, better they get ordered to become Nationalized, all their IP licensed GNU, a budget allocated for their maintenance, data harvesting stripped, and anyone can submit PRs.
True public infrastructure.
Most corporate-facing businesses have some sense of sanity because they know every purchase of their product is going to be scrutinized to hell by a team of MBA accountants. They’ve got to actually meet expectations versus consumer business which can be sold (and defended) on marketing And hype alone
I mean, a lot of internal corporate “AI” systems I’ve seen are ChatGPT wrappers, and total messes. But they get approval and funding.
Another odd thing is that IBM’s hardly ‘selling’ here. These are basically open source contributions for PR, with the hope that they’re used internally or in their other businesses I guess. But it’s quite low-key for PR.
They also made IBM Watson which was not a commercial success. They’re already familiar with ‘overpromise and underdeliver’ nature of AI.