- nazgul17The RealLifeLore YouTube channel published a video about this not long ago: https://youtu.be/n8kSGH4I8Ps
- Thinking aloud, but couldn't someone create a website with some malicious text that, when quoted in a prompt, convinces the LLM to expose certain private data to the web page, and couldn't the webpage send that data to a third party, without the need for the LLM to do so?
This is probably possible to mitigate, but I fear what people more creative, motivated and technically adept could come up with.
- Can't you say the same of the human brain, given a different algorithm? Granted, we don't know the algorithm, but nothing in the laws of physics implies we couldn't simulate it on a computer. Aren't we all programs taking analog inputs and spitting actions? I don't think what you presented is a good argument for LLMs not "know"ing, in some meaning of the word.
- Not so sure. The government has placed a A$50M incentive per violation discovered, I heard. That sounds like a powerful incentive on the companies to outsmart the kids.
- That's not an interesting difference, from my point of view. The box m black box we all use is non deterministic, period. Doesn't matter where on the inside the system stops being deterministic: if I hit the black box twice, I get two different replies. And that doesn't even matter, which you also said.
The more important property is that, unlike compilers, type checkers, linters, verifiers and tests, the output is unreliable. It comes with no guarantees.
One could be pedantic and argue that bugs affect all of the above. Or that cosmic rays make everything unreliable. Or that people are non deterministic. All true, but the rate of failure, measured in orders of magnitude, is vastly different.
- I feel like you're putting words in someone else's mouth. Maybe you are not responding to OP but, in your mind, to an ex-colleague that did so in a different venue than this forum?
In a forum like this, stating your preference is just that: stating your preference.
If you were talking with your manager and stated your preference, you'd be stating your preference and, between the lines, asking to make it happen for yourself.
If you were talking with your manager and stated your preference and specified the reason is because you prefer working around people, only then, between the lines, you'd be asking to make it happen for your whole team.
- Probably, because not everyone is made happy: some are annoyed. I am not going to enter the merit of either side.
- To add to this, the system offering text generation, i.e. the loop that builds the response one token at a time generated by a LLM (and at the same time feeds the LLM the text generated so far) is a Markov Model, where the transition matrix is replaced by the LLM, and the state space is the space of all texts.
- They also tried to heal the damage, to partial avail. Besides, it's science: you need to test your hypotheses empirically. Also, to draw attention to the issue among researchers, performing a study and sharing your results is possibly the best way.
- Brexit was fairly recent, compared to the bad governance of Argentina.
- The scalability of spying has exploded. Back before re-election comms, the government had no way to spy on communications and sieve out opposers - now they do, with encryption the only thing standing in the way.
- We should rename it as BrowserScript. The .bs extension is a funny bonus.
- In my humble (backender) opinion, if it's hard to use a tool right, that counts as a cons, and that must be accounted for when choosing which tool to use.
- What a condescending and arrogant answer.
I think you should also think more about this.
One can believe that Apple (or any company) should let you do whatever you want with your hardware - in general - and point out any instance when they don't; even if that specific instance is not something that touches you!
This is true of everything. Another example: if you believe in freedom of speech, you should vocally defend anyone who is deprived of it, even when that is not you. Otherwise, you lose by divide and conquer.
Apes together strong.
- Both can be true
- I worked on a Spring Boot application that would take more than a minute to start. Not Spring's fault, IIUC, but nevertheless a focus killer.
- This is true, but I find my train of thought slips away if I have to wait more than a handful of seconds, let alone two minutes.
Tying this back to your point, those limited hours of focus time come in blocks, in my experience, and focus time is not easily "entered", either.
- You put my thoughts into words better than I could.
This reminds me of the exploration-exploration trade-off in reinforcement learning: you want to maximise your long term profits but, since your knowledge is incomplete, you must acquire new knowledge, which companies do by trying stuff. Prematurely dismissing GenAI could mean missing out on new efficiencies, which take time to be identified.
- That's the dream. Reality is very different. Mypy presents numerous false negatives and false positives. Useful to screen for some bugs, but definitely far from giving guarantees.
Not to mention, if a library does not or does sloppily use type annotations, you would not get reliability even with a perfect type checker.
- Except it's not in alpha phase