When my next cell phone costs $2500, how will it be Joe Biden’s fault?
Posts by Kendall Clark
Silence from the but-her-emails set is deafening only if you think they actually believed a word of it. The did not.
I’ve never once had a feeling of inadequacy in the face of AI outputs, for two reasons, I suppose: (1) it never does anything truly surprisingly good; (2) we made it: our glorious works are our glory.
nmn.gl/blog/agi-gods
Yes and yr bitcoin stash won’t save you.
Forgot AGI; I’d be happy if ChatGPT could produce ascii-only text so I can cut and paste manually to email myself.
Like a fucking caveman!
Really focused on the big issues here. And it’s a fun list of flaws but it assumes something that is far from obvious, the AI Feet of Clay thesis. If a model fails at small things, then it’s worthless, not intelligent, etc.
We don’t apply that standard to people. Consider the absent-minded genius.
www.informationweek.com/machine-lear...
@ejimenez-ruiz.bsky.social Can you add me to the KG starter pack? Founder & CEO, Stardog. Chaired SPARQL WG, edited SPARQL Protocol spec, etc.
Bruises that won’t heal.
*approximate
Alinc cancelled my order after a month of silence. Meh.
And yeah maybe he’s not yr guy; but the only other sane candidate as best prose stylist of American English in the 19th century was born a slave!
Reminding you of the extraordinarily weird fact that the best prose stylist of American English—maybe of any language, actually—of the 19th century—maybe of any century, actually—was the American president.
The actual fuck was going on back then!?
Reading Labatut’s The Manaic and I’m struck by the fact that in the 20th century Turing, von Neumann, and Gödel were all alive at the same time.
What other century could boast half as much or even a third?
Followup: Pensando 10/25 SmartNICs are plentiful on eBay. But the 200gbe are not. Colfax Direct is often the move here, especially since it reps Xilinx which, like Pensando, is part of AMD now. Alinc.com had DSC-200 and we’ll see how they are to work with.
Unobtanium, apparently: Pensando DSC-200, Bluefield-3, Qualcomm AI 100, Chelsio T7, etc. it’s not just GPUs.
The pursuit of AGI has led to one-big-model blindness.
The number of LLMs will appreciate M x N x P, where M is cognitive jobs-to-be-done, N is industry verticals, and P is significant, non-overlapping firms.
In other words, there will be a LOT of models, soon and then forever. #llm
Also keyboards (in my 10th year of this obsession), DPUs, HCI, Proxmox, AI appliance design, startups.