It's Machine Learning, yo

[Mongrel]

Between image compositing, chatbots and other stuff, lots of this shit has been in the news this year and last. So... dedicated thread?


Also, Clarkesworld, one of Sci-Fi/Fantasy's few remaining literary magazines, and also one of, if not the fastest and most efficient publisher for responding to story submissions (also one of the best-paying), just had to close all submissions due to being flooded with chatbot-esque submissions.

EDIT: Meanwhile our old pal Sharkey is having fun. But who is Scott Sharkey, I mean, really?

[Thad]

There's some debate as to whether or not machine learning is a proper subset of AI or merely a closely-related and overlapping field that grew out of the study of AI.

There's an argument to be made that "machine learning" is a better descriptive phrase for engines like ChatGPT, especially given that "artificial intelligence" is one of those terms that means one thing in academia but has a very different connotation with laypeople, largely due to its extensive history in science fiction.

Or you could just insult people's intelligence because they know more than you; that's certainly a rich and storied tradition, too.

I'm partial to "spicy autocomplete", myself.

[Mongrel]

I'm pretty far from being the sharpest tool in the shed, but if I'm dumber than academia as a whole, we're in even worse trouble than I thought. >:V

I will tweak the title, though perhaps not in the manner you'd prefer.

I'm partial to "spicy autocomplete", myself.

Okay, that's a good one.

[Thad]

I mean I gotta hand it to you, making an assertion that's confidently wrong, then responding to a correction by feeding the same statement through a thesaurus and then repeating it? I can't think of a more fitting way to kick off a discussion about the current generation of Spicy Autocomplete tools.

Perfect; no notes.

[Mongrel]

Thad, sometimes I think you just enjoy being angry just a little too much.

In any case, you and I and academia are all going to be stuck with the public conflating "AI" with AGI and maybe that just becoming a messy confusing umbrella term anyway, because that's language for ya.

EDIT: Actually I'm enjoying the way the forum cuts off the title on the homepage, so changing it to that.

[Brantly B.]

It kinda looks like you're intentionally needling people on the political board fyi.

[Mongrel]

It kinda looks like you're intentionally needling people on the political board fyi.

If you or Thad or anyone else here for that matter posted a thread with "you idiots" in the title, do you think it would it be assumed either of you were directing that at the boards?

Or do you just mean me snarking at Thad after he clucked about the title?

[Mongrel]

I'll extend the same offer here I was gonna PM Thad: Why don't we just delete everything in here after the initial post so the thread can just exist for its intended purpose?

[Khan97]

There's some debate as to whether or not machine learning is a proper subset of AI or merely a closely-related and overlapping field that grew out of the study of AI.

There's an argument to be made that "machine learning" is a better descriptive phrase for engines like ChatGPT, especially given that "artificial intelligence" is one of those terms that means one thing in academia but has a very different connotation with laypeople, largely due to its extensive history in science fiction.

Or you could just insult people's intelligence because they know more than you; that's certainly a rich and storied tradition, too.

I'm partial to "spicy autocomplete", myself.

Hello,

It is true that there is some debate about the relationship between machine learning and artificial intelligence. While machine learning is a subfield of AI, some argue that it is distinct enough to be considered an entity unto itself. However, it's worth noting that machine learning is often used as a tool within the broader field of AI, such as chatbots, here to learn more.

When it comes to using language to describe AI and machine learning, there are certainly differing views on the most accurate or appropriate terminology. While "artificial intelligence" has a certain cachet in popular culture, it is important to use terminology that is clear and accessible to a broader audience. "Machine learning" can certainly be a useful and more descriptive term in some contexts.

The field of AI and machine learning is complex and constantly evolving, and there is always more to learn and discuss.

[sei]

Do we not use the strong/weak or general/narrow distinctions for AI any more?

I think laymen see "AI" and perhaps mistakenly interpret it as in the sense of general AI, but that doesn't necessarily mean narrow AI systems aren't some form of what academia means by "AI."

[Mongrel]

Do we not use the strong/weak or general/narrow distinctions for AI any more?

I think laymen see "AI" and perhaps mistakenly interpret it as in the sense of general AI, but that doesn't necessarily mean weak AI systems aren't some form of what academia means by "AI."

I think General/Narrowed are useful terms here, but really anything that lays out a simple, clear distinction which the public can grok between AGI and everything else would be good.

[zaratustra]

There's some debate as to whether or not machine learning is a proper subset of AI or merely a closely-related and overlapping field that grew out of the study of AI.

The issue is that the discipline of AI is entirely composed of closely-related and overlapping fields.

[Upthorn]

Machine learning is a necessary but insufficient condition of Artificial Intelligence.

[Niku]

are we just going to ignore that mongrel got an actual ai to come in here and machine learn

[Caithness]

Looked more like machine-splaining to me

[sei]

Machine learning is a necessary but insufficient condition of Artificial Intelligence.

Expert systems fall under AI and don't need to involve ML.

[Upthorn]

I would contend that Expert systems are also a necessary but insufficient condition of Artificial intelligence.

[Upthorn]

I would contend that a true artificial intelligence must have
Some way to use known rules to make decisions (expert systems)
Some way to remember the outcomes of previous decisions (machine learning)
Some way to recognize patterns within those outcomes to modify known rules and conceive previously unknown rules (I don't know if this is even a field)

[Mongrel]

The most fascinating thing to me so far is that Chat GPT, while in no way self-aware, has been trained on human content (and it's worth talking about what content and how, and who trained it), and has manifested all the buried paranoid insecurity of HAL 9000.

Arthur C. Clarke would have laughed at how accidentally prescient he was - not in predicting how an actual AGI would act, but in having such a well-tuned instinct for what sort of logical and illogical processes our purpose-built machines would absorb from us most readily.

[Mazian]

There's a wild case happening right now in a federal court in New York. It seems to have started as a fairly mundane injury case, until one of the lawyers decided to use ChatGPT to come up with citations to support their brief. Opposing council noticed that these citations did not seem to exist, and convinced the judge to demand that the first lawyer provide copies of those decisions... so in response, they doubled down and also used ChatGPT to generate those.

Someone is about to become a cautionary tale.

Case documents are fun: the weirdness starts in document 24, is fully entrenched by document 30, and the attachment to document 32 includes some actual ChatGPT queries used by the lawyer in question, like "are the other cases you provided fake".