ChatGPT ranks the 2010-2020 drafts

[Dead Coyote]

I like how Lafreniere is listed in two different drafts

 

[Cizin]

I like how Lafreniere is listed in two different drafts

and Boldy is listed twice in the same draft (as Matt and Matthew)

 

[3074326]

Wait until it decides over your employment and insurances.

It already is, in some cases at least.

The job I had in 2013 was testing software that filtered out applicants based on arbitrary personality tests..... and that was a music store (instruments). They relied on it for applicants shortly after. They also just went out of business. Good riddance.

Anyways, just looking at the first draft, Brett Connolly should be way higher than a lot of those guys. It's actually insane that it has him that low. Hard to take it seriously with that kind of weird pick. He has 5x more goals than Beau Bennett, 2x more assists, and 2.5x games. He was a solid player. Bennett never cracked 20 points in a season and is ranked 17.

 

[kingsholygrail]

I'm no longer worried about AI.

 

[Jumptheshark]

Any reason why it seems to have forgot Taylor Hall for 2010?

so it not just me

 

[rattle]

and Boldy is listed twice in the same draft (as Matt and Matthew)

and they are a few tiers apart !

 

[Golden_Jet]

Straight up missing Kaprizov in 2015, apparently the universe just decided he was never going to be a first rounder.

Marner didn’t make the first round either.

 

[MikeyMike01]

When will people understand that so-called AI can’t actually think in any capacity, and can do nothing more than imitate human language? You might as well ask a parrot.

 

[AvroArrow]

Any reason why it seems to have forgot Taylor Hall for 2010?

He's probably a conservative, chatGPT isn't actually openAI, it's programmed to create things based on a pre-determined set of views

 

[JelloPuddyPops]

Hahaha... Kole Lind #27 and no Hischier.

Yeah, I think we are safe from SkyNet for a while.

Okaayyy, ChatGPT, time to put down the bottle of ripple. Now create us a picture of humans with seven fingers on each hand.

 

[Soundwave]

Taylor Hall falling from no.1 to not even on the list is certainly a look.

 

[Mr Kot]

When will people understand that so-called AI can’t actually think in any capacity, and can do nothing more than imitate human language? You might as well ask a parrot.

This. A.I is basically "pay no attention to that man behind the curtain" level

 

[JTToilinginToronto]

Better than a lot of the takes regularly seen on HFBoards

 

[BlueSeal]

Wait until it decides over your employment and insurances.

No one's paying attention to that until it's far too late.

 

[Enniskillen]

Soulless computer included comments about Rodion Amirov, good that John Connor put an end to this madness.

 

[Sam de Mtl]

Glad to see Ty Smilanic is now a 1st rounder in a redraft. We were just about to release his rights, but maybe we should offer him max ELC instead.

 

[Breakers]

Definately taking Michael Dal Colle over thatcher demko

ChatGPT really is next level genius.

 

[ijuka]

When will people understand that so-called AI can’t actually think in any capacity, and can do nothing more than imitate human language? You might as well ask a parrot.

You're sort of selling it short. It's akin to google except it can generate answers to questions that people might not have answered before, and can at least point you in the right direction.

Also, it's far superior to humans in complex mechanical problems, especially ones that can trick the human brain's assumptions.

Something like ChatGPT generates its answers by using random noise, which means that its answer to the same question is probably different every single time. And yeah, it's not as useful for specialized purposes, because it's very general and censored.

Take these draft rankings, for example. They're not great. But what if you had created a model specifically to rank NHL players? It would likely perform far better.

 

[66-30-33]

Once AI gets better (ASI maybe?) it will be scary good at player evaluations. Scouts? just plug in ASI.

 

[Esq]

Well, when the comments on HF are fed into its model, what do you expect?

 

[TheDawnOfANewTage]

Huh, tried the same experiment but it just keeps asking me where I’d place John Conner. Is that a prospect or somethin?

 

[Joe MacMillan]

You're sort of selling it short. It's akin to google except it can generate answers to questions that people might not have answered before, and can at least point you in the right direction.

Also, it's far superior to humans in complex mechanical problems, especially ones that can trick the human brain's assumptions.

Something like ChatGPT generates its answers by using random noise, which means that its answer to the same question is probably different every single time. And yeah, it's not as useful for specialized purposes, because it's very general and censored.

Take these draft rankings, for example. They're not great. But what if you had created a model specifically to rank NHL players? It would likely perform far better.

I highly doubt that. It constantly struggles with computing mathematical expressions correctly (even the most trivial ones) and rarely provides valuable insight into the implementation of source code for computer programs. Both of these are very formal and mechanical tasks.

It's an LLM, not true AI. The only thing it's good for is semantic analysis and text generation, and is more advanced at these tasks than most humans.

 

[Romang67]

Complaining about ChatGPT's inability to do rankings on hockey players is like trying to screw a nail with a saw and then complaining about the results.

 

[MadLuke]

inability to do rankings on hockey players

According to people reaction everytime humans do one in the media, it could be just something impossible to do.

 

[Romang67]

According to people reaction everytime humans do one in the media, it could be just something impossible to do.

One could even say that human inability to do consistent rankings with solid underlying reasoning is why the generative AI is so bad at it.

GIGO.