Comment by jjcon | Hacker News Reader

Comment by jjcon | Hacker News Reader

Parent Full thread

> AI is neither artificial nor intelligent. It is made from natural resources and it is people who are performing the tasks to make the systems appear autonomous.

People performing the tasks? As in coding tasks? That’s kinda the definition of artificial. If AI spawned naturally… that’s what wouldn’t be artificial.

Or is she referring to data labeling and forgetting the many non supervised areas of AI? (Not that labeling would make it less artificial)

Or is she suggesting that anything made from natural resources is natural… is anything not made from natural resources? The reason for artificial in artificial intelligence is to juxtapose against biological intelligence not speak to resource usage.

I think arguing on the ‘not intelligent’ front is fine but that’s just kind of a word game. The field seeks intelligence and is fine making stupid ai if it means better ai later. Unless she is getting into notions of a soul or some bs about consciousness in which case we have left the realm of science altogether.

Any way you slice it seems more like an inane point for making a headline than one for substantive discussion.

It seems to me somewhat similar to an argument put forward by Jaron Lanier: that what is termed artificial intelligence’s (or ‘deep learning’’s etc.) current successes, at least as far as something like the translation services of the big corporations for example, is actually built by leveraging corpuses that are the fruits of a huge amount of individual human effort. Sometimes, say with the ‘Mechanical Turk’ thing that Amazon has, or CAPTCHAs or something, this is a more explicit connection (…bit Wizard of Oz! :) As I understand it, he proposes that an alternative to UBI etc. might be micro-compensations (or transactions) in return for providing this data. This might be a prelude/transition to a stage where our basic needs (on a sort of Mazlow’s hierarchy) were met by A.I., or that there might be increasingly creative or interesting ways to complete some tasks that we could go on refining forever.

> are the fruits of a huge amount of individual efforts

That doesn’t make it any less artificial does it? If anything that makes it more so. That is also again only speaking to supervised learning (and only the fraction of it where datasets are curated not collected) not AI in general

I agree - it’s kind of lazily framed/worded. A.I. is a vast field after all?

That's because that was an inane point used as a headline that didn't really have much to do with what the article is actually about.

It's a criticism of the training data sets used on ai's that were manually tagged by individuals that's led to some extreme biases in ai behaviour resulting in real world consequences.

For a great real world example of this happening right now, there's a fairly large scandal and a whole bunch of angry people over the problems caused by latitude's choice of training material for their ai driven text adventure game.

https://gitgud.io/AuroraPurgatio/aurorapurgatio

Glancing at the table of contents...

> X. Count Grey: Enslavement and Slaughter of Children

>XI. Kyros: Kidnapping, Enslavement, Unethical Experimentation, and Torture

What the hell am I about to read?

The training data for Ai Dungeon.

https://play.aidungeon.io/

> that didn't really have much to do with what the article is actually about.

That was a quote of the author not just the headline

> It's a criticism of the training data sets used on ai's that were manually tagged by individuals

Again though this is just a fraction of AI. Lumping all of AI into the realm of ‘manually curated supervised learning’ is just incorrect, yet the author seems to think that’s all there is, it’s either that or they are just trying to make a headline.

> For a great real world example of this happening right now

I hardly think a controversy about a random text game on the internet warrants such a serious tone

To me she is referring to what she says previously:

> Also, systems might seem automated but when we pull away the curtain we see large amounts of low paid labour, everything from crowd work categorising data to the never-ending toil of shuffling Amazon boxes.

So the artificial and intelligent part is a tiny piece of the task, most of it is done through humans and logistics.

Again though, ”manual curation of data for supervised learning tasks” is just a fraction of the ai research being done today. Making these claims as to apply to all AI is either ignorant or sensationalist regardless of the truthfulness of the claim itself.

Reply on news.ycombinator.com