Well, I asked Chat GPT the following query:

Tell me how to win at no limit texas hold em.

This was it's response:


Texas Hold'em is a complex and dynamic game that requires a combination of skill, strategy, and luck to win. Here are a few tips that may help you improve your chances of winning at No Limit Texas Hold'em:

Know the Rules and Hand Rankings: Familiarize yourself with the rules of the game and the hand rankings, as this will help you make informed decisions during the game.

Study Player Behavior: Pay attention to how other players behave, including their tendencies and weaknesses. This can help you make better decisions and adjust your strategy accordingly.

Manage Your Bankroll: Proper bankroll management is crucial in No Limit Texas Hold'em, as it helps you avoid going broke and allows you to make smart decisions at the table.

Position is Key: Position is important in Texas Hold'em, as it can give you an advantage by allowing you to act last on subsequent rounds.

Don't Play Too Many Hands: Don't play too many hands, especially those that are not strong. Focus on playing only premium hands and be patient.

Bluffing: Bluffing can be an effective tool in No Limit Texas Hold'em, but use it wisely and sparingly. Know when to bluff and when to fold.

Know When to Fold: Don't be afraid to fold if you have a weak hand or if the pot odds are not in your favor. This can help you conserve your chips and make better decisions later in the game.

Practice: The more you play, the better you will become. Take advantage of free online games and practice as much as you can to improve your skills.

Remember, winning at No Limit Texas Hold'em requires a combination of skill, strategy, and luck. By following these tips and continually learning and practicing, you may be able to increase your chances of success at the table.

Scary. It could probably crunch all the data in 20 years of two plus two posts and come up with good advice for individual hands.

Not inconveivable it's already done so as well as articles written. Raises a ton of IP ownership/credit questions.

Of course it may over-index on GGs posts, and it will be reccomending everyone limp their entire range from Early Poistion.

Speaking as a mathematician who works with data for a living, I can tell you that ChatGPT is trained to sound smart, not to be smart. It can’t actually do the calculations required to learn the game. When mathematicians ask it math questions and look for accuracy, it is comically wrong. There is a reason there are stories about how it can pass exams from law schools and business schools but not math or science departments.

yup, it told me all continuous functions are differentiable, which is complete nonsense.

Lol

Yeah, I asked it to give a proof that there exists a real number that is not the solution to any algebraic equation and it produced a very well-written argument that was complete gibberish.

I suspect it’s partly an issue that ChatGPT hasn’t been trained on enough mathematics textbooks.

At the same time, I am doubtful these large language models will in the near future be able to create any new mathematics that fall outside the scope of their training data.

I would think learning math would be easier for AI than learning language. If chatgpt isn't good at math yet, it has to be simply because that has not been its focus.

GTO is pretty much done. It's just a matter of putting it into words.

Not really. The AI can memorize the statement of a theorem but that doesn’t mean it understands how to apply the theorem to solve examples.

ChatGPT definitely knows the definition of a linear subspace. But when I ask it if y=x+1 is a linear subspace of the plane, it falsely says “yes” and gives a gibberish proof. It just isn’t very good at applying theorems/definitions to specific examples. It knows things but it can’t correctly apply mathematical reasoning.

One issue is that mathematics has the property that making very small changes to a statement (i.e modifying a hypothesis) can change a true statement to a false statement and vice versa. That is not true of English essays where almost everything is an opinion, and the task is to write convincingly rather than correctly.

It is not just Mathematics where ChatGPT struggles. I asked it some basic questions, literally sophomore level stuff, in my primary field (ChemEng). It gives a very convincing but completely wrong answer multiple times.

I also ask it a simple question in a different Engr field it made a complete bonehead mistake. It’s answer was completely self contradictory. When I pointed out it’s error, it saw and agreed to the mistake, changed its answer.

That error was incredibly simple. A confident and observant middle schooler could have seen it. Basically asked about a theoretical and practical efficiency. ChatGPT put the practical max efficiency above the theoretical max. Given the numbers it shared I could make an educated guess how and why it made the error. But the key is it simply did not understand the source materials it was referencing.