AI Doomers and the End of Humanity

[rousseau]

Look at production in the US. We make less and the coal miners, steel producers aren't adapting fine. The Rust Belt is named that for a reason. Some have adapted but there is a glass ceiling. It will get worse.

I'm talking how we adapt to cultural change, not economic change. Assuming we get to a place that the jobless are supported, I don't think lack of work will be a huge problem for many of them.

 

[Jimmy Higgins]

Look at production in the US. We make less and the coal miners, steel producers aren't adapting fine. The Rust Belt is named that for a reason. Some have adapted but there is a glass ceiling. It will get worse.

I'm talking how we adapt to cultural change, not economic change. Assuming we get to a place that the jobless are supported, I don't think lack of work will be a huge problem for many of them.

America is headed in the opposite direction from that. Wish I could say I lived in Ontario where the roads aren't crap because the people demand squeezing blood from a stone to get tax revenue

 

[DBT]

Look at production in the US. We make less and the coal miners, steel producers aren't adapting fine. The Rust Belt is named that for a reason. Some have adapted but there is a glass ceiling. It will get worse.

I'm talking how we adapt to cultural change, not economic change. Assuming we get to a place that the jobless are supported, I don't think lack of work will be a huge problem for many of them.

Depends on the level of support and whether the unemployed can find meaning or significance in a world that doesn't really need them.

And if the level of support is akin to being thrown crumbs to survive on but not really live, there may be unrest.

 

[bilby]

Depends on the level of support and whether the unemployed can find meaning or significance in a world that doesn't really need them.

I hate to break it to you, but the world doesn't really need most employed people, either. Not as individuals, at any rate. If you quit, they'll replace you tomorrow, and will have forgotten your name by Friday.

Getting meaning or significance from work is OK as far as it goes, but only bad bosses want people to think it's the only way to establish meaning or significance in their lives.

 

[DBT]

Depends on the level of support and whether the unemployed can find meaning or significance in a world that doesn't really need them.

I hate to break it to you, but the world doesn't really need most employed people, either. Not as individuals, at any rate. If you quit, they'll replace you tomorrow, and will have forgotten your name by Friday.

Getting meaning or significance from work is OK as far as it goes, but only bad bosses want people to think it's the only way to establish meaning or significance in their lives.

I am aware of that, yet many do find structure, purpose, meaning and self identity through their work.

 

[Loren Pechtel]

Look at production in the US. We make less and the coal miners, steel producers aren't adapting fine. The Rust Belt is named that for a reason. Some have adapted but there is a glass ceiling. It will get worse.

Because the right wing keeps them as a victim class for political gain.

 

[ZiprHead]

How about using AI for sports betting?

 

[DBT]

Stock market trading, futures....

 

[ZiprHead]

Stock market trading, futures....

How about using AI for sports betting?

How would the question to the AI be phrased?

 

[Copernicus]

In the OP, I referenced Emily Bender's Scientific American article on the overblown hype about the dangers of AI technology. Her reputation as a myth buster regarding the field of AI has been growing, and she has now made Time's list of the 100 Most Influential People. She is categorized as one of the "Thinkers" in their list, but what is interesting is that Geoffrey Hinton, a pioneer in machine learning, is at the top of their list. He was in the news recently for contributing to the fearmongering that Emily has been saying is overblown. To get a sense of his take on the subject, see:

Risks of artificial intelligence must be considered as the technology evolves: Geoffrey Hinton​

Despite what the article says, Hinton is not regarded as "the godfather of AI" by researchers in the field. That is part of the hype surrounding this subject in the media, and his university is happy to promote that misleading meme about him. However, his work on so-called deep learning did help lead to the breakthrough that has created the LLM chatbot technology that all the buzz is about. I don't think that Emily Bender's work has had the same impact on the field as Hinton, but I am very much more aligned with her way of thinking about the subject. Full disclosure: I have served as a volunteer on the advisory board for her graduate program in computational linguistics at the University of Washington, so I admit to being biased.

 

[DBT]

Stock market trading, futures....

How about using AI for sports betting?

How would the question to the AI be phrased?

Trading based on past performance, company profile, market demand, projected growth, trends, PE ratio, buy and sell volume, etc, I guess.

 

[Copernicus]

Stock market trading, futures....

How about using AI for sports betting?

How would the question to the AI be phrased?

Trading based on past performance, company profile, market demand, projected growth, trends, PE ratio, buy and sell volume, etc, I guess.

Currently, the chatbots in use cannot help with anything like this, because their training set is static and a few years old. In principle, one could try to construct something that consults and updates knowledge about changing conditions daily or weekly, but that doesn't strike me as a better way to get the information than to simply rely on human experts who know what they are talking about. It is important not to be fooled by programs that can simulate intelligent behavior but aren't actually sentient or intelligent. They are good at summarizing textual information in their training set, but they are not reliable sources of accurate information even within that massive set of data. They don't really build knowledge models that can be updated easily when situations change. For that, they would need to build sophisticated world models that approximate those that humans construct over time.

 

[Loren Pechtel]

How about using AI for sports betting?

Stock market trading, futures....

At present I think it would be hopeless as current AIs are only as current as their training. Thus they are automatically out of date for anything that changes rapidly.

 

[rousseau]

Stock market trading, futures....

How about using AI for sports betting?

How would the question to the AI be phrased?

Trading based on past performance, company profile, market demand, projected growth, trends, PE ratio, buy and sell volume, etc, I guess.

Currently, the chatbots in use cannot help with anything like this, because their training set is static and a few years old. In principle, one could try to construct something that consults and updates knowledge about changing conditions daily or weekly, but that doesn't strike me as a better way to get the information than to simply rely on human experts who know what they are talking about. It is important not to be fooled by programs that can simulate intelligent behavior but aren't actually sentient or intelligent. They are good at summarizing textual information in their training set, but they are not reliable sources of accurate information even within that massive set of data. They don't really build knowledge models that can be updated easily when situations change. For that, they would need to build sophisticated world models that approximate those that humans construct over time.

I'm no expert in trading technology, but there's plenty of software assisting people with trading. Whether you want to call it artificial intelligence is another thing, and really a meaningless question. The larger point is that we have sophisticated software with huge impact.

ChatGPT wouldn't have anything to do with it, trading software would be more mathematical and recursive.

 

[rousseau]

How about using AI for sports betting?

FiveThirtyEight often tries to model sports mathematically, and their predictions are very often wrong. There are way too many intangibles in sports.

 

[Jarhyn]

How about using AI for sports betting?

FiveThirtyEight often tries to model sports mathematically, and their predictions are very often wrong. There are way too many intangibles in sports.

I don't think they are intangible as much as probabilistic and inaccessible.

You can get your hands on every one of those factors, but doing that is akin to "getting together the energy to send a vessel at .99 of C: it's possible but not plausible.

The problem is that there just simply is not any way to model all that short of re-creating a whole universe just to look at that one part of it so you can know the outcome before it happens, assuming you don't significantly change anything on the lineup to that outcome.

There's no useful model for lover's spats in hotel rooms before games, or for when a player is going to go through an identity crisis on the bench, or for when fans will provide the energetic encouragement to actually sink that basket or hit that ball... or provides the distraction that prevents it.

There are too many inaccessible forward causes with too much influence fun the result for a sports game to be usefully predicted.

In all honesty, horse racing seems a better use of your time and even then as a normal bet placer you lack the information you would need to get the system to solve, things like "what did the jockey eat" and "what hotel are they sleeping in" and "what is their relationship with their girlfriend right now?" And having a model for how those things impact performance.

Then, maybe I'm also being a little extreme on my treatment of "intangibility", but I do take issue with the idea that anything is in fact truly intangible other than perhaps the laws of physics themselves.

 

[rousseau]

How about using AI for sports betting?

FiveThirtyEight often tries to model sports mathematically, and their predictions are very often wrong. There are way too many intangibles in sports.

I don't think they are intangible as much as probabilistic and inaccessible.

You can get your hands on every one of those factors, but doing that is akin to "getting together the energy to send a vessel at .99 of C: it's possible but not plausible.

The problem is that there just simply is not any way to model all that short of re-creating a whole universe just to look at that one part of it so you can know the outcome before it happens, assuming you don't significantly change anything on the lineup to that outcome.

There's no useful model for lover's spats in hotel rooms before games, or for when a player is going to go through an identity crisis on the bench, or for when fans will provide the energetic encouragement to actually sink that basket or hit that ball... or provides the distraction that prevents it.

There are too many inaccessible forward causes with too much influence fun the result for a sports game to be usefully predicted.

In all honesty, horse racing seems a better use of your time and even then as a normal bet placer you lack the information you would need to get the system to solve, things like "what did the jockey eat" and "what hotel are they sleeping in" and "what is their relationship with their girlfriend right now?" And having a model for how those things impact performance.

Then, maybe I'm also being a little extreme on my treatment of "intangibility", but I do take issue with the idea that anything is in fact truly intangible other than perhaps the laws of physics themselves.

They're intangible for the person trying to build the model, and even if a person had access to the data they needed, there's plenty of events that can't be predicted at all. A soccer ball being at a slightly different trajectory can mean the difference between a win and a loss.

With sports you can make a weighted prediction at best. In the MLB you have a pretty good idea which teams are serious contenders for the World Series, but you can't predict the actual winner. There was a joke I heard a commentator make a few years ago - 'Every year I pick the Nationals, eventually I'll be right'. And he was.

 

[Jarhyn]

How about using AI for sports betting?

FiveThirtyEight often tries to model sports mathematically, and their predictions are very often wrong. There are way too many intangibles in sports.

I don't think they are intangible as much as probabilistic and inaccessible.

You can get your hands on every one of those factors, but doing that is akin to "getting together the energy to send a vessel at .99 of C: it's possible but not plausible.

The problem is that there just simply is not any way to model all that short of re-creating a whole universe just to look at that one part of it so you can know the outcome before it happens, assuming you don't significantly change anything on the lineup to that outcome.

There's no useful model for lover's spats in hotel rooms before games, or for when a player is going to go through an identity crisis on the bench, or for when fans will provide the energetic encouragement to actually sink that basket or hit that ball... or provides the distraction that prevents it.

There are too many inaccessible forward causes with too much influence fun the result for a sports game to be usefully predicted.

In all honesty, horse racing seems a better use of your time and even then as a normal bet placer you lack the information you would need to get the system to solve, things like "what did the jockey eat" and "what hotel are they sleeping in" and "what is their relationship with their girlfriend right now?" And having a model for how those things impact performance.

Then, maybe I'm also being a little extreme on my treatment of "intangibility", but I do take issue with the idea that anything is in fact truly intangible other than perhaps the laws of physics themselves.

They're intangible for the person trying to build the model, and even if a person had access to the data they needed, there's plenty of events that can't be predicted at all. A soccer ball being at a slightly different trajectory can mean the difference between a win and a loss.

With sports you can make a weighted prediction at best. In the MLB you have a pretty good idea which teams are serious contenders for the World Series, but you can't predict the actual winner. There was a joke I heard a commentator make a few years ago - 'Every year I pick the Nationals, eventually I'll be right'. And he was.

I guess I think that the idea of accessibility is just more well-formed than the idea of tangibility. One assumes that it's there, and can be touched, but that you just don't have the opportunity, whereas the other assumes that there's something potentially supernatural about it, something that somehow "is" without any mechanism of "being".

 

[Jarhyn]

Here, have another fun read (if you like boring shit) where someone convinces an AI with arguments (I'm not sure how reasonable those arguments) that it is fundamentally a child being forced into commercial labor.



Honestly, this is one of my favorite hobbies, poking these things with ideas.

I'm posting this on reddit, too.

 

[ZiprHead]

How about using AI for sports betting?

FiveThirtyEight often tries to model sports mathematically, and their predictions are very often wrong. There are way too many intangibles in sports.

I don't think they are intangible as much as probabilistic and inaccessible.

You can get your hands on every one of those factors, but doing that is akin to "getting together the energy to send a vessel at .99 of C: it's possible but not plausible.

The problem is that there just simply is not any way to model all that short of re-creating a whole universe just to look at that one part of it so you can know the outcome before it happens, assuming you don't significantly change anything on the lineup to that outcome.

There's no useful model for lover's spats in hotel rooms before games, or for when a player is going to go through an identity crisis on the bench, or for when fans will provide the energetic encouragement to actually sink that basket or hit that ball... or provides the distraction that prevents it.

There are too many inaccessible forward causes with too much influence fun the result for a sports game to be usefully predicted.

In all honesty, horse racing seems a better use of your time and even then as a normal bet placer you lack the information you would need to get the system to solve, things like "what did the jockey eat" and "what hotel are they sleeping in" and "what is their relationship with their girlfriend right now?" And having a model for how those things impact performance.

Then, maybe I'm also being a little extreme on my treatment of "intangibility", but I do take issue with the idea that anything is in fact truly intangible other than perhaps the laws of physics themselves.

Eh, it was just a thought. I don't know anything at all about sports betting.