this post was submitted on 21 Oct 2024
932 points (97.3% liked)

Technology

60052 readers
2774 users here now

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


Our Rules


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

Approved Bots


founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
[–] [email protected] 33 points 2 months ago (3 children)

10 to 30? Yeah I think it might be a lot longer than that.

Somehow everyone keeps glossing over the fact that you have to have enormous amounts of highly curated data to feed the trainer in order to develop a model.

Curating data for general purposes is incredibly difficult. The big medical research universities have been working on it for at least a decade, and the tools they have developed, while cool, are only useful as tools too a doctor that has learned how to use them. They can speed diagnostics up, they can improve patient outcome. But they cannot replace anything in the medical setting.

The AI we have is like fancy signal processing at best

[–] [email protected] 16 points 2 months ago (1 children)

Not an expert so I might be wrong, but as far as I understand it, those specialised tools you describe are not even AI. It is all machine learning. Maybe to the end user it doesn't matter, but people have this idea of an intelligent machine when its more like brute force information feeding into a model system.

[–] [email protected] 24 points 2 months ago (1 children)

Don't say AI when you mean AGI.

By definition AI (artificial intelligence) is any algorithm by which a computer system automatically adapts to and learns from its input. That definition also covers conventional algorithms that aren't even based on neural nets. Machine learning is a subset of that.

AGI (artifical general intelligence) is the thing you see in movies, people project into their LLM responses and what's driving this bubble. It is the final goal, and means a system being able to perform everything a human can on at least human level. Pretty much all the actual experts agree we're a far shot from such a system.

[–] [email protected] -3 points 2 months ago (3 children)

It may be too late on this front, but don't say AI when there isn't any I to it.

Of course it could be successfully argued that humans (or at least a large amount of them) are also missing the I, and are just spitting out the words that are expected of them based on the words that have been ingrained in them.

[–] [email protected] 8 points 2 months ago

This is not up to you or me : AI is an area of expertise / a scientific field with a precise definition. Large, but well defined.

[–] [email protected] 5 points 2 months ago* (last edited 2 months ago) (1 children)

AI as a field of computer science is mostly about pushing computers to do things they weren't good at before. Recognizing colored blocks in an image was AI until someone figured out a good way to do it. Playing chess at grandmaster levels was AI until someone figured out how to do it.

Along the way, it created a lot of really important tools. Things like optimizing compilers, virtual memory, and runtime environments. The way computers work today was built off of a lot of things out of the old MIT CSAIL labs. Saying "there's no I to this AI" is an insult to their work.

[–] [email protected] 2 points 2 months ago (1 children)

Recognizing colored blocks in an image was AI until someone figured out a good way to do it. Playing chess at grandmaster levels was AI until someone figured out how to do it.

You make it sound like these systems stopped being AI the moment they actually succeeded at what they were designed to do. When you play chess against a computer it's AI you're playing against.

[–] [email protected] 0 points 2 months ago* (last edited 2 months ago) (1 children)

That's exactly what I'm getting at. AI is about pushing the boundary. Once the boundary is crossed, it's not AI anymore.

Those chess engines don't play like human players. If you were to look at how they determine things, you might conclude they're not intelligent at all by the same metrics that you're dismissing ChatGPT. But at this point, they are almost impossible for humans to beat.

[–] [email protected] 2 points 2 months ago

I'm not the person you originally replied to. At no point have I dismissed ChatGPT.

I disagree with your logic about the definition of AI. Intelligence is the ability to acquire, understand, and use knowledge. A chess-playing AI can see the board, understand the ramifications of each move, and respond to how the pieces are moved. That makes it intelligent - narrowly so, but intelligent nonetheless. And since it’s artificial too, it fits the definition of AI.

[–] [email protected] 2 points 2 months ago

Intelligence: The ability to acquire, understand, and use knowledge.

A self-driving car is able to observe its surroundings, identify objects and change its behaviour accordingly. Thus a self-driving car is intelligent. What's driving such car? AI.

You're free to disagree with how other people define words but then don't take part in their discussions expecting everyone to agree with your definiton.

[–] [email protected] 1 points 2 months ago (1 children)

AI in health and medtech has been around and in the field for ages. However, two persistent challenges make roll out slow-- and they're not going anywhere because of the stakes at hand.

The first is just straight regulatory. Regulators don't have a very good or very consistent working framework to apply to to these technologies, but that's in part due to how vast the field is in terms of application. The second is somewhat related to the first but really is also very market driven, and that is the issue of explainability of outputs. Regulators generally want it of course, but also customers (i.e., doctors) don't just want predictions/detections, but want and need to understand why a model "thinks" what it does. Doing that in a way that does not itself require significant training in the data and computer science underlying the particular model and architecture is often pretty damned hard.

I think it's an enormous oversimplification to say modern AI is just "fancy signal processing" unless all inference, including that done by humans, is also just signal processing. Modern AI applies rules it is given, explicitly or by virtue of complex pattern identification, to inputs to produce outputs according to those "given" rules. Now, what no current AI can really do is synthesize new rules uncoupled from the act of pattern matching. Effectively, a priori reasoning is still out of scope for the most part, but the reality is that that simply is not necessary for an enormous portion of the value proposition of "AI" to be realized.

[–] [email protected] 1 points 2 months ago

The oversimplification was intended - you also caught my meaning of it being able to synthesize new rules.