Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Gemini "duck" demo was not done in realtime or with voice

twitter.com

Post reply on HN
Print view View original

631–640 of 683 posts

Re: Gemini "duck" demo was not done in realtime or with voice

#631

Post by nullc »

JohnFen wrote: post #624
10u152 wrote: post #432

Earlier quoted context omitted.

Related, I work with industrial control systems. We’d call this “smoke and mirrors”. Sometimes the client would insist on seeing a small portion of a large project working well before it was ready. They’d misunderstand that 90% of the bulk of the work is not visible to the user but they’d want to see a finished state. We’d set up a dummy HMI and have someone pressing buttons on it for the demo, and someone in the nex…

I, too, work with industrial control systems. If any of us did that sort of thing, we'd be fired instantly -- and rightfully so.

There would be no problem if you told the client that you were faking the back end behavior and if the client's motivation is that they wanted to see the workflow to make sure that there wasn't a misunderstanding on what you were supposed to be implementing, then a mocked backend would be perfectly fine for their purposes.

Re: Gemini "duck" demo was not done in realtime or with voice

#632

Post by nullc »

thewakalix wrote: post #440
gardenhedge wrote: post #436

Earlier quoted context omitted.

Is this cheating? It sounds like cheating and reflects quite poorly on you.

> It was our first comp sci class ever, we were given raspberry pi's. We had no coding experience or guidance, and were asked to create "something". Garbage in, garbage out.

It's plausible to me that they weren't provided with what they needed precisely because pervasive cheating allowed their predecessor classmates to complete the assignments.

Re: Gemini "duck" demo was not done in realtime or with voice

#633

Post by JohnFen »

nullc wrote: post #631
JohnFen wrote: post #624

Earlier quoted context omitted.

I, too, work with industrial control systems. If any of us did that sort of thing, we'd be fired instantly -- and rightfully so.

There would be no problem if you told the client that you were faking the back end behavior and if the client's motivation is that they wanted to see the workflow to make sure that there wasn't a misunderstanding on what you were supposed to be implementing, then a mocked backend would be perfectly fine for their purposes.

Absolutely true. The comment, however, very, very strongly implied that the client wasn't made aware that the demo was faked. If the client was made aware, then why would they have someone hiding in the next room to "make it seem like it was working"?

Re: Gemini "duck" demo was not done in realtime or with voice

#634

Post by rkeene2 »

freejazz wrote: post #628
rkeene2 wrote: post #529

Earlier quoted context omitted.

It's not clear because it is false and I believe I can produce a proof if you are willing to validate that you accept my premise. Your CPU, right now, has known defects. It will produce the wrong outputs for some inputs. It seems to meet your definition of broken. Do you agree with that premise ?

One has nothing to do with the other. There's no rule about all broken tools because they can be broken in different ways. What's so difficult about my hypothetical? I laid it all out for you.

I assumed you understood we reached the end of the usefulness of your hypothetical to the original analogy since, as you said, the tools can be broken in different ways. I tried to introduce a scenario that was more applicable and less theoretical so that we could discuss those particular points.

If we do somehow try to apply your analogy, it would indicate that the LLM output is flawed in a way we cannot scrutinize -- the hidden failures that we aren't detecting (why? It's not specified, I am assuming because we didn't check to see if the tool was "broken" and not meeting some unspecified quality-level; that is, it's an unknown unknown failure mode).

This doesn't really comport with the LLM scenario, where the output is fully viewed, and the outputs are widely understood (that is it is a known failure mode).

This is more closely related to a computing service -- of which you are an active user of. You are using a "broken" computer right now, according to your definition of broken correct ?

Re: Gemini "duck" demo was not done in realtime or with voice

#635

Post by plaidfuji »

zmmmmm wrote: post #561
plaidfuji wrote: post #337

Earlier quoted context omitted.

These LLMs do not have a concept of factual correctness and are not trained/optimized as such. I find it laughable that people expect these things to act like quiz bots - this misunderstands the nature of a generative LLM entirely. It simply spits out whatever output sequence it feels is most likely to occur after your input sequence. How it defines “most likely” is the subject of much research, but to optimize for f…

> because for certain prompts, the approximate consensus of all available text on the internet is pretty much true I think you're slightly mischaracterising things here. It has potential to be at least slightly and possibly much better than that. This is evidenced by the fact it is much better than chance at answering "novel" questions that don't have a direct source in the training data. Why it can do it is because…

You’re absolutely right and I’m sure that something resembling higher-level pattern matching is present in the architecture and weights of the model, I’m just saying that I’m not aware of “logical thought” being explicitly optimized or designed for - it’s more of a sometimes-emergent feature of a machine that tries to approximate the content of the internet, which for some topics is dominated by mostly logical thought. I’m also unaware of a ground truth against which “correct facts” could even be trained for..

Re: Gemini "duck" demo was not done in realtime or with voice

#636

Post by plaidfuji »

eurekin wrote: post #354
plaidfuji wrote: post #337

Earlier quoted context omitted.

These LLMs do not have a concept of factual correctness and are not trained/optimized as such. I find it laughable that people expect these things to act like quiz bots - this misunderstands the nature of a generative LLM entirely. It simply spits out whatever output sequence it feels is most likely to occur after your input sequence. How it defines “most likely” is the subject of much research, but to optimize for f…

The first question I always ask myself in such cases: how much input data has a simple "I don't know" lines? This is clearly a concept (not knowing sth) that has to be learned in order to be expressed in the output.

Ha, probably an insignificant amount. The internet is nothing if not confidently-stated positive results, no matter how wrong they might be. No wonder this is how LLMs act.

Re: Gemini "duck" demo was not done in realtime or with voice

#637

Post by qingcharles »

nar001 wrote: post #527
qingcharles wrote: post #494

Earlier quoted context omitted.

I demo'd a full browser office suite in 1998 called Office Wherever (o-w.com). It used Java applets to do a lot of the more tricky functions. Shopped it around VCs. Got laughed out of all the meetings. "Companies storing their documents on the Internet?! You're out of your mind!"

Im curious if the code would be available somewhere? I have to admit I'm curious how it worked!

Long lost! Unless Internet Archive has a copy of the site? Never checked! The main domain was officewherever.com.

Re: Gemini "duck" demo was not done in realtime or with voice

#638

Post by adriand »

TerrifiedMouse wrote: post #550
eloisius wrote: post #512

Earlier quoted context omitted.

It’s hard even when you don’t take into consideration that you don’t know what you’ve misremembered. Try writing a memoir. You’ll realize you never actually remember what anyone says, but caveating your dialog with “and then I think they said something like” would make horrible reading.

Don’t people just bulk qualify the dialogue, e.g. “I don’t remember the exact words. Tom said … then Dick replied … something like that”. Often we don’t quote people and instead provide a high level description of what they said, e.g. “Harry described the problems with his car.”, where detail is omitted.

Sure. Some memoirs state this in the foreword. There are also memoirs that are "fictionalized", like James Frey's book "A Million Little Pieces." Originally published as a memoir in 2003, it was later revealed that many of the events Frey described were exaggerated or fabricated. This caused a lot of controversy at the time, but many subsequent memoirs followed this pattern and I think it's become quite accepted in the genre.

Re: Gemini "duck" demo was not done in realtime or with voice

#640

Post by gardenhedge »

frail_figure wrote: post #569
gardenhedge wrote: post #436

Earlier quoted context omitted.

Is this cheating? It sounds like cheating and reflects quite poorly on you.

The view up on that high horse must be interesting! Were you the kid who reminded teachers about homework? Literally all that matters is that they passed.

> Were you the kid who reminded teachers about homework?

Are you trying to bully me or something? Not going to work with me. You've revealed your poor character with that comment.

Post reply on HN