bsk7
bsk7 on Hacker News
> An inaccurate engine and a statistical one are indistinguishable.
How so? The paper is trying to assess if some semantic understanding is being created by the underlying model, not if a completely accurate one is. If such a world model maps to Othello-prime (gleaning concepts like tiles, colors etc), that is still a very interesting result.
> We only need a single counter example to show that Othello-GPT does not have a systemic understanding of the rules, only a statistical inference of them.
I don't agree with this binary interpretation. This only indicates that any systemic understanding the model has internally built is not complete. We are not trying to assess completeness of that understanding. E.g. if you had a traditional rules engine for Othello, and you removed one rule that could result in illegal moves, does that make that rule engine a statistical model all of a sudden?
Exactly. Consider adding a link to your LinkedIn profile or your CV on your Medium post so folks that are looking to help (referrals, hiring managers, recruiters) can review your skills and experience.
To me, the bigger question is, why do these interviews require every candidate to prove their programming/design skills from scratch? Say I have worked at a respected tech company for N years, my record there should provide a lot more signal than a day’s worth of interviewing would.
I think we need some sort of standardized metrics for different dimensions of tech skills, and those metrics update over years of experience. Then companies don’t need to assess tech skills - they just need to look at your record, knowing that it is objective. Certifications, open-source contributions are some examples. The problem is the work you do closed-door in a company cannot be objectified or standardized easily. If we can solve that, that would go a long way here.
Interviews can then focus on soft and behavioral skills solely. Not to say that those interviews are any more useful or signal-deducing as they happen today.
I totally agree. Especially for "micro-interactions" between people while remaining anonymous (i.e. not needing to exchange phone numbers, email addresses etc.).
E.g.:
- receiving appointment info at a doctor's office (in lieu of a physical appointment card)
- exchanging links/addresses/numbers with strangers in any kind of conversational interaction ("call this number", "go to this address", "here's a brochure" etc.) without divulging personal identifiers
Phone-to-phone, anonymous, exchange-and-forget type interactions. Can use bluetooth, wifi, or data connections (or even just qrcodes). I've always thought that there would be usefulness for something like this (key being simplicity and anonymity).