polotics

4 ms·

31.
by polotics 1mo ago
yep "next-embedding" predictor is more correct, and not just at the end but through the layers, and folding back dimensions into that one next token is one small final step, and next-embedding could be named "next-meaning&quo
32.
by polotics 1mo ago
Not trying to be provocative or anything, but ELIZA 60 years ago was already a better psychoanalist than many people I know. `By your metrics, we achieved AGI in 1966.` https://en.wikipedia.org/wiki/ELIZA
33.
by polotics 1mo ago
Nah as long as you don't give them first-names, I think you should be ok. also: Pretty pleaser please people don't get your agents registered as direct-reports in the org-chart with HR!
34.
by polotics 1mo ago
Call me when Astra gets the Nobel prize... We'll have AGI when prizes have two categories, one for assisted humans, and one for pure-AI.
35.
by polotics 1mo ago
At the end of the first paragraph there is this sentence: "Training support is a hypothesis consistent with these results, not something this study proves." I understand it, but I find the wording very unnatural. To improve readab
36.
by polotics 1mo ago
OK so there is a better angle to look at all this, there really is: https://www.researchgate.net/publication/357868645_Edgar_All...
37.
by polotics 1mo ago
This is so weird, because using a script to invoke the agy cli is clearly using a "third party tool to access the service" !
38.
by polotics 1mo ago
What is preventing anyone from automating on top of the normal google Antigravity CLI?
39.
by polotics 1mo ago
I disagree. This trolling and misdirection are not in the movie. Also, both are clearly effective, and a discourse that warps economic impact on US midterms voters within identities, reduces the chance of exposing the bad consequences and r
40.
by polotics 1mo ago
More detail on the Patrick Boyle video where the actual picture of Bessent's note that included that purchase is mentioned. Also the fact that these are specifically French bonds that were dumped with no warning. This is dropping the m
41.
by polotics 1mo ago
well in my experience mostly you get either: - stupidity - stupidity and malice
42.
by polotics 1mo ago
I'm still very impressed by qubes, and glad I'm not such a target that I feel I need the level of opsec it affords (on all my laptops). Maybe someday AI-assisted killchains will be so widespread that Qubes is the minimal level for
43.
by polotics 1mo ago
please refrain from referring to countries as if they were individuals
44.
by polotics 1mo ago
The Savior fallacy. With power more concentrated, the ability to avoid being impacted for those in power rises and rises and rises. How are incentives then aligned? I'm sure that having 10 million governments would be much better than
45.
by polotics 1mo ago
Make your own benchmark on your own work, keep them to yourself. Try one typical not completely unambiguous spec document like you're likely to have seen. See if the AI asks the right questions, and how it navigates its unknown unknown
46.
by polotics 1mo ago
I kind of agree that the post is a bit too far away from the trenches to be able to claim it will reveal "software engineering reality". From where I snipe, I see a big divide between those... 1) that try to surrender to AI, aimin
47.
by polotics 1mo ago
obesity and food-supply are not in a linear relationship, pumping out some high sucrose corn syrup is one thing, having all the nutrients the population needs to stay healthy is another
48.
by polotics 1mo ago
This is hilarious, and so true!
49.
by polotics 1mo ago
Wow, that's a different perspective. I should read more of your posts. For me the fact that 30% of men and 20% of women are not having intimacy is quite scary. I guess a revised edition of the book sextinction is in order.
50.
by polotics 1mo ago
Well, also screwdrivers can do that... if you actually do use AI to do some proper research on a fields you don't really know that well, looking for peer reviewed papers, actually figuring more things out thanks to AI, then I don'
51.
by polotics 1mo ago
if your agent's software delivers real world value to some people, that's motivation enough I find
52.
by polotics 1mo ago
I find that actually asking the LLM what they've done and why, and pointing out everything that's not quite right, is a good way to avoid that kind of overload. if you're actually managing to have fifteen simultaneous work tr
53.
by polotics 1mo ago
Did you really mean to write really: "Online girlfriends are cool" In my humble opinion, they are the epitome of uncool.
54.
by polotics 1mo ago
Can you define your use of the word 'smartest' here just in case some of us don't quite know what you mean?
55.
by polotics 1mo ago
Have we really though? This is not a silly question, because it's kind of silly: Has an AI ever actually gone through all of the steps necessary to become a chess master? Go to competitions, raise in the ranks, take a plane to the mast
56.
by polotics 1mo ago
this is software development
57.
by polotics 1mo ago
So exactly true! With ideas people unbridled, the painting-yourself-in-a-corner tendency that was always there has gone exponential, add to it the not-completely-aligned incentives, and the oh-so-clever impressive writer of long prose LLMs
58.
by polotics 1mo ago
Yep and I just ran a simple enterprisey "ambiguity" bench on the big three (US) model providers: same ambiguous initial-prompt with same clarifications and pushback prompt sequence afterwards. The edge of correct/better when
59.
by polotics 1mo ago
This quote from the post makes me sad for all the researchers and all the math and the scaling and the investing that I believe was done in good faith in pursuit of pure science and betterment for us all: "AI is designed to make you de
60.
by polotics 1mo ago
Hi. This is very interesting, could you link to the research? There is a dearth of proper research studies that A/B test what approach is best in terms of harness structure based on repeatable benchmark data with relevant sample uses-c

More ›