Technology

59422 readers

3001 users here now

This is a most excellent place for technology news and articles.

Our Rules

Follow the lemmy.world rules.
Only tech related content.
Be excellent to each another!
Mod approved content bots can post up to 10 articles per day.
Threads asking for personal tech support may be deleted.
Politics threads may be removed.
No memes allowed as posts, OK to post as comments.
Only approved bots from the list below, to ask if your bot can be added please contact us.
Check for duplicates before posting, duplicates may be removed

Approved Bots

founded 1 year ago

MODERATORS

131

Chatbot letdown: Hype hits rocky reality (www.axios.com)

submitted 7 months ago by Telodzrum@lemmy.world to c/technology@lemmy.world

51 comments fedilink hide all child comments

Grumbles about generative AI's shortcomings are coalescing into a "trough of disillusionment" after a year and a half of hype about ChatGPT and other bots.
Why it matters: AI is still changing the world — but improving and integrating the technology is raising harder and more complex questions than first envisioned, and no chatbot has the magic answers.
Driving the news: The hurdles are everything from embarrassing errors, such as extra fingers or Black founding fathers in generated images, to significant concerns about intellectual property infringement, cost, environmental impact and other issues.

you are viewing a single comment's thread
view the rest of the comments

[–] FaceDeer@fedia.io 2 points 7 months ago* (last edited 7 months ago) (1 children)

You've switched from saying that a cryptocurrency's "primary purpose" is as a currency for transactions to saying that they're securities, those are not remotely similar things.

Anyway. Are you aware that, assuming the Gartner hype cycle actually does apply here (it's not universal) and AI is really in the "trough of disillusionment", beyond that phase lies the "slope of enlightenment" wherein the technology settles into long-term usage? I feel like you're tossing terminology around in this discussion without knowing a lot about what it actually means.

No, it can't, because it isn't and cannot be made trustworthy. If you need a human to review the output for hallucinations then you might as well save yourself the licensing costs and let the human do the work in the first place.

If you think it can't replace anyone then why say "It can't replace anyone senior"?

Also, what licensing costs? Some AI providers charge service fees for using them, but as far as I'm aware none of them claim copyright over the output of LLMs. And there are open-weight LLMs you can run yourself on your own computer if you want complete independence.

[–] knightly@pawb.social 0 points 7 months ago (1 children)

You've switched from saying that a cryptocurrency's "primary purpose" is as a currency for transactions

That was someone else's argument, but it can't be denied that the original use-case envisioned blockchain securities as a currency.

Anyway. Are you aware that, assuming the Gartner hype cycle actually does apply here (it's not universal) and AI is really in the "trough of disillusionment", beyond that phase lies the "slope of enlightenment" wherein the technology settles into long-term usage? I feel like you're tossing terminology around in this discussion without knowing a lot about what it actually means.

I'm well aware, it's the slope down into the trough that will pop the bubble of "AI" overinvestment. I wouldn't be surprised if some application of LLM tech finds a profitable niche, but I'd be very surprised to see it in common use outside of automated copywriting for scammers.

No, it can't, because it isn't and cannot be made trustworthy. If you need a human to review the output for hallucinations then you might as well save yourself the licensing costs and let the human do the work in the first place.

If you think it can't replace anyone then why say "It can't replace anyone senior"?

That wasn't me.

Also, what licensing costs? Some AI providers charge service fees for using them,

Those ones.

but as far as I'm aware none of them claim copyright over the output of LLMs.

Hasn't it already been ruled that LLM outputs cannot be copyrighted, or was that just patents and I'm misremembering?

And there are open-weight LLMs you can run yourself on your own computer if you want complete independence.

Ah yes, because rolling your own unreliable text generator is so much less expensive. XD

[–] FaceDeer@fedia.io 0 points 7 months ago

That wasn't me.

Apologies, you're right. I took care to double check the wording but neglected to spot the different username.

Hasn't it already been ruled that LLM outputs cannot be copyrighted, or was that just patents and I'm misremembering?

No, there have been a lot of misleading news articles with headlines like that but nothing like that has been decided in any jurisdictions that I'm aware of.

The most popular news story to get headlines like that is the Thaler v. Perlmutter case, if you do a Google search for that you'll find an endless stream of "U.S. Court holds that AI generated works cannot be copyrighted" headlines. But that's not remotely what the case was actually about. What actually happened was that Thaler generated an image using an AI and then went to the US Copyright office to register the copyright *in the name of the program that generated it." That is, he went to the Copyright office and told them "the copyright for this work is solely held by the computer that generated it. Nobody else was involved in its creation." The copyright office responded "that's silly, copyright can only be held by a person (human or corporate). A computer is not a person." Since the list of copyright-holders Thaler was claiming was therefore zero, the Copyright office ruled that the work must be in the public domain.

Thaler sued, and in the subsequent court case he tried to add himself to the list of copyright-holders. The judge said "no, that's not what this suit is about, knock it off. You told the copyright office you didn't hold a copyright to that work, and as a result their ruling that the work was uncopyrighted was correct."

If he'd tried to claim copyright for himself from the start there wouldn't have been any problem. There have been other instances where humans have registered copyrights for works that they used an AI to generate. The only reason Thaler failed was because he specifically and explicitly said that he wasn't claiming copyright over it. This has unfortunately turned into one of those "suing McDonalds for making their coffee hot" semi-urban-legends.

And even if a U.S. court did make a ruling along those lines, the U.S. isn't the whole world. There are plenty of countries out there that would be happy to take the lead instead if the U.S. decided it didn't want to be supportive of local AI-driven industry.

Ah yes, because rolling your own unreliable text generator is so much less expensive. XD

It really is. I run LLMs on my home computer myself, for fun, using a commodity graphics card. The models it can run don't quite reach ChatGPT's level of sophistication but they're close, and they have the advantage that I can control them much more precisely to perform the tasks that I want them to perform. If I wanted to use a more sophisticated open model there are cloud providers that could run it for me for pennies, I just like having the hardware completely under my control.