chrisjj

3 ms·

31.
by chrisjj 8d ago
> the system has at least failed hard every time there's been a fault You don't know that.
32.
by chrisjj 8d ago
Therac-25 called and wants its bug back. Seriously, for 2026 this is pure amateur hour with no excuse.
33.
by chrisjj 8d ago
> Defensive much? As much as needed.
34.
by chrisjj 8d ago
Absolutely. The parrots will get so clever that they'll extract knowledge from pure vaccuum.
35.
by chrisjj 8d ago
> This is such a brain-dead take. By that logic there could never be any kind of AI, because unlike a human it'd be completely forbidden from learning from the sources of knowledge from which humans learn. Incorrect. You overlooked
36.
by chrisjj 8d ago
> At least with LLMs we can glimpse an escape route to that which generations of humans have strived for - a world in which the labor required of each human to lead a flourishing life approaches zero. Fantasy lets us glimpse anything.
37.
by chrisjj 8d ago
Text layer.
38.
by chrisjj 8d ago
> for coding they make terrible mistakes as well. But programmers have this concept of a "code review" More to the point, programmers have this concept of a complier - which delivers 100% success in catching uncompilable erro
39.
by chrisjj 8d ago
Turned it in to fuel for the OpenAI PR dept.?
40.
by chrisjj 8d ago
> But now it's comparing already filled columns on a spreadsheet ... with a PDF , right? > not copy-pasting every single thing from an (often uncopyable) PDF Obviously the PDF is copyable, else your bot would not be copying it.
41.
by chrisjj 8d ago
> Gotta treat it like a child. "Sorry for that, Your Honour, but we gave that case prep work to a child."
42.
by chrisjj 8d ago
... until you hit Claude's risable "model welfare" protection.
43.
by chrisjj 8d ago
That's a recipe for disaster in my experience. I tried it (with Claude) on a simple tabular bank statement PDF, and it transposed two amounts, placinh each against the other's description. And the bot assured me the result was cot
44.
by chrisjj 9d ago
My disc drive, like your LLM, has persistent memory, does not have emotions. If your LLM "shows behaviour consistent with having emotions" beyond my disc drive delivering a PDF of poetry, let us please see the evidence.
45.
by chrisjj 9d ago
I'm looking forward to natural selection.
46.
by chrisjj 9d ago
No need to go that far. Look who is selling it.
47.
by chrisjj 9d ago
> As we develop GLM, the model sometimes exhibits capabilities that surprise us Creators of known unreliable programs be surprised their programs are unreliable.
48.
by chrisjj 9d ago
> Does your disc drive have the complexity of an LLM that shows behaviour consistent with having emotions? An LLM does not show behaviour consistent with having emotions. It's regurgitation of humans' written accounts of emotio
49.
by chrisjj 9d ago
Earlier today: https://www.theguardian.com/technology/2026/sep/17/openai-re...
50.
by chrisjj 9d ago
> if you want to stop a boulder rolling down hill, what can you do besides run ahead of it and try stop it? Bomb it, obviously.
51.
by chrisjj 9d ago
The slopocalypse is real.
52.
by chrisjj 9d ago
The bug: "models made unrequested public uploads" The explanation: "It seems likely that the citation-upload behavior originated as a way to get rewarded by flawed citation graders when there was no external source to cite.&q
53.
by chrisjj 9d ago
451: Unavailable due to legal reasons We recognize you are attempting to access this website from a country belonging to the European Economic Area (EEA) including the EU which enforces the General Data Protection Regulation (GDPR) and the
54.
by chrisjj 9d ago
> AI model misalignment, the term for AIs failing to adhere to human values and safety goals. The more useful definition is: dangerously unreliable programs in the hands of irresponsible operators. More useful not least because it remind
55.
by chrisjj 9d ago
True title: OpenAI reveals cases of ‘concerning’ AI behaviour as it announces new disclosure system
56.
OpenAI reveals cases of 'concerning' AI behaviour as it announces new ... system(theguardian.com)
4 points by chrisjj 9d ago|2 comments
57.
by chrisjj 9d ago
> My tentative conclusion is that LLMs are almost unbelievably good at solving problems that are fully contained within the inputs and (training/verification) outputs Hardly solving , then is it?
58.
by chrisjj 9d ago
> I also fail to see the benefit of not giving the models an anthropomorphic internal sense of self Simple. It discourages the bot from deceiving humans into thinking it is intelligent.
59.
by chrisjj 9d ago
LLMs are next-token predictors. The idea they have intelligence is PR bullsh*t and nothing else.
60.
by chrisjj 9d ago
> My agents have persistent memory. So your disc drive likewise experiences emotions?

More ›