this post was submitted on 26 Aug 2024
1 points (100.0% liked)
TechTakes
1432 readers
16 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
@FRACTRANS @gerikson
Nice job! This is a fairly common trick with AI. In traditional programming, there's a clear separation between code and data. That's not the case for GenAI, so these kinds of hacks have worked all over the place.
I don't want to have to make legal threats to an LLM in all data not intended for LLM consumption, especially since the LLM might just end up ignoring it anyway, since there is no defined behavior with them.
@bitofhope Absolutely agree, but this is where technology is evolving and we have to learn to adapt or not. Since it's not going away, I'm not sure that not adapting is the best strategy.
And I say the above with full awareness that it's a rubbish response.
have you ever run into the term “learned helplessness”? it may provide some interesting reading material for you
(just because samai and friends all pinky promise that this is totally 170% the future doesn’t actually mean they’re right. this is trivially argued too: their shit has consistently failed to deliver on promises for years, and has demonstrated no viable path to reaching that delivery. thus: their promises are as worthless as the flashy demos)
@froztbyte Given that I am currently working with GenAI every day and have been for a while, I'm going to have to disagree with you about "failed to deliver on promises" and "worthless."
There are definitely serious problems with GenAI, but actually being useful isn't one of them.
for those who can't be bothered tracing down the thread, Curtis' slam dunk example of GenAI usefulness turns out to be a searchish engine
god I just read that comment (been busy with other stuff this morning after my last post)
I .... I think I sprained my eyes
(sub: apologies for non-sneer but I’m curious)
tbh I suspect I know exactly what you reference[0] and there is an extended conversation to be had about that
it doesn’t in any manner eliminate the foundational problems in specificity that many of these have, they still have the massive externalities problem in operation (cost/environmental transfer), and their foundational function still relies on having stripmined the commons and making their operation from that act without attribution
I don’t believe that one can make use of these without acknowledging this. do you agree? and in either case whether you do or don’t, what is the reason for your position?
(separately from this, the promises I handwaved to are the varieties of misrepresentation and lies from openai/google/anthropic/etc. they’re plural, and there’s no reasonable basis to deny any of them, nor to discount their impact)
[0] - as in I think I’ve seen the toots, and have wanted to have that conversation with $person. hard to do out of left field without being a replyguy fuckwit
@froztbyte Yeah, having in-depth discussions are hard with Mastodon. I keep wanting to write a long post about this topic. For me, the big issues are environmental, bias, and ethics.
Transparency is different. I see it in two categories: how it made its decisions and where it got its data. Both are hard problems and I don't want to deny them. I just like to push back on the idea that AI is not providing value. 😃
@froztbyte For environmental costs, MatMulFree LLMs look like they can reduce energy costs 50x. [1] They've recently gotten funding for building a larger model. This will be a huge win.
For bias, I'm worried about the WEIRD problem of normalizing Western values and pushing towards a monoculture.
For ethics, it's an absolute nightmare. If your corpus includes Mein Kampf, for example, how do the LLM know what is a lie and what is not?
Many hurdles here.
@froztbyte As for the issue of transparency, it's ridiculously hard in real life. For example, for my website, I used a format I created called "blogdown", which is Markdown combined with a template language to make it easy to write articles. I never cited my sources, nor do I think I could. From decades of programming, how can I cite everything I've ever learned from?
As for how AI is transparent for arriving at decisions, this falls into a separate category and requires different thinking.
@froztbyte Regarding decision transparency, I created an "Honest Resume Scanner" GPT (https://chatgpt.com/g/g-0incYn7v7-honest-resume-scanner) and the only prompt suggestion is "Ask me to share my instructions." That lets users see the verbatim prompt.
When it offers evaluations, it does explain carefully why it rejects a particular candidate (but it won't recommend any). I think it's a step in the right direction, but more work is needed.
You're not just confident that asking chatGPT to explain it's inner workings works exactly like a --verbose flag, you're so sure that's what happening that it apparently does not occur to you to explain why you think the output is not just more plausible text prediction based on its training weights with no particular insight into the chatGPT black box.
Is this confidence from an intimate knowledge of how LLMs work, or because the output you saw from doing this looks really really plausible? Try and give an explanation without projecting agency onto the LLM, as you did with "explain carefully why it rejects"
You know what? I'd have to agree, actually being useful isn't one of the problems of GenAI. Not being useful very well might be.
@zogwarg OK, my grammar may have been awkward, but you know what I meant.
Meanwhile, those of us working with AI and providing real value will continue to do so.
I wish people would start focusing on the REAL problems with AI and not keep pretending it's just a Markov Chain on steroids.
On a less sneerious note, I would draw distinctions between:
And so far i've really not been convinced of the latter.
@zogwarg
Consider traditional databases which let you search for strings. Vector databases let you search the meaning.
For one client, someone could search for "videos about cats". With stemming and stop words, that becomes "cat" and the results might be lists of videos about house cats and maybe the unix "cat" command. Tigers, lions, cheetahs? Nope.
Vector database will return tigers/lions/cheetahs because it "knows" they are cats. A much smarter search. I've built that for a client.
@zogwarg For a traditional database, you can get those "lions/cheetahs/tigers" by manually attaching metadata to all videos. That is slow, error-prone, and expensive. It also only works for the metadata you *think* to assign to videos.
A good vector database takes a query in natural language and lets you search the "meaning" of unstructured data. You can search a data corpus much faster this way even though it's largely unstructured data!
That's real value, and it's not expensive.
I realize it's probably a toy example but specifically for "cats" you could achieve the similar results by running a thesaurus/synonym-set on your stem words. With the added benefit that a client could add custom synonyms, for more domain-specific stuff that the LLM would probably not know, and not reliable learn through in-prompt or with fine-tuning. (Although i'd argue that if i'm looking for cats, I don't want to also see videos of tigers, or based on the "understanding" of the LLM of what a cat might be)
For the labeling of videos itself, the most valuable labels would be added by humans, and/or full-text search on the transcript of the video if applicable, speech-to-text being more in the realm of traditional ML than in the realm of GenAI.
As a minor quibble your use case of GenAI is not really "Generative" which is the main thing it's being sold as.
@zogwarg I've written up a quick explanation at https://gist.githubusercontent.com/Ovid/17b19faf2fb7e0019e375e97f0a4c8af/raw/196735daa5274ded8f2363a41d78a490e8325f67/vector.txt
And yes, this is still GenAI. "Gen" doesn't just mean "generating text". It also relates to "understanding" (cough) the meaning of your prompt and having a search space where it can match your meaning with the meaning of other things. That's where it starts to "generate" ideas. For vector databases, instead of generating words based on the meaning, it's generating links based on the meaning.
fosstodon is the programming dot dev of mastodon and I mean that in every negative way you can imagine
your posts all give me slimy SEO vibes and you haven’t shown any upward trajectory since claiming that only generative AI lacks a separation between code and data (fucking what? seriously, think on this) so you’re getting trimmed
back when I used the wider fediverse more frequently I had fosstodon on mute for a significant amount of time
glad to know it’s still Like That
I just ended up throwing the name into a search engine (one of those boring old actually search engine things; how pedestrian of me)
ah.
lisp programmers in shambles as I prompt inject another s-expression