Anyone been playing with gpt? I see the chat bot is trending now
Write a commentary on a Go/Baduk/Weiqi tournament, including Shin Jinseo, Ke Jie, Go Seigen and Magnus Carlsen.
The Go/Baduk/Weiqi tournament held recently was an exciting event for fans of the game. It featured some of the world’s greatest players, including Shin Jinseo, Ke Jie, Go Seigen, and Magnus Carlsen. Shin Jinseo is a young Korean prodigy and the current world champion. His skill in the game is extraordinary and he has made a name for himself as a formidable opponent. Ke Jie is the former world champion, and is known for his creative and unpredictable moves. Go Seigen is a living legend in the game, having been a professional for over 70 years, and his experience and skill are unparalleled. Finally, Magnus Carlsen is a Norwegian grandmaster who is known for his out-of-the-box thinking and calculated strategies.
The tournament was an intense battle of wits and strategy. Shin Jinseo emerged victorious, displaying a mastery of the game that impressed even the other players. Ke Jie put up a valiant effort, but was unable to overcome Shin’s impressive skill. Go Seigen’s skill and experience were on full display, and it was a privilege to witness such a master in action. Magnus Carlsen, while not victorious, showed that his skill in the game is still formidable.
The tournament was an exciting and thrilling event, and a great showcase of the skill and talent of these four players. Each one demonstrated their unique style and strategies, and it was a pleasure to watch these masters of the game compete.
It’s still astonishing to me how these chat bots can simultaneously be extremely knowledgable while also lacking almost any form of understanding. It’s a bit like the essays a high school student might write about a book they hadn’t read, based on some summary found online. The whole text makes sense, but it’s clear that the student / AI has no idea what they’re talking about.
we may overestimate level of understanding of humans
so, bots will reach human level sooner than we expect
not because bots will become smart
but because we are more stupid than we think.
But that’s an empty statement. Lowering the bar by calling humans stupid does not change that GTP convincingly seems to believe there is no problem with the inclusion of Magnus Carlsen in a professional Go tournament. The AI knows Carlsen plays chess, but can’t connect those simple dots. Basically, the AI will be just as happy writing an essay on why 2+2=4 as it is on writing an essay on 2+2=5.
And I indeed believe bots will become more intelligent than humans in the near future, but there’s still a long road to go for that. On the other hand, AI research seems to progress almost exponentially, so I’m not going to wager putting an estimate on it.
humans can write science fiction or comedy fantasy about world where 2+2=5
GPT don’t know what you wish from him - to find error in your question or to write novel about other world.
I seen prompt somewhere after which GPT-3 able to say which question is correct and which is not instead of answering these questions.
From my perspective there is an interesting question behind this ‘stupidity comparison’. I’ve heard AI sceptics call all the recent improvements ‘glorified parameter fitting’ and I guess they are right. OTOH, who knows, maybe what we call human intelligence and conscience may in the end be correctly described as glorified synaptic strength fitting …
I think you value GPT far too highly. From what I have seen it is riddled with inconsistencies and when pointing them out it evades responsibility by claiming it knew these were inconsistent. That’s all that’s needed to fool people into believing it’s smart, but I won’t consider an AI smart until it can actually be used for things that require understanding.
Smart humans invent new things, write books worth reading, do novel research and expand the boundaries of what we know. AI can’t do that yet, so although it is a very impressive knowledge base, it doesn’t appear to be “smart” to me yet.
(and I would honestly not be too surprised if we reach the point where such AI do exist within the next 5 years, at least I find that more realistic than that we do not reach such a point for another 50 years)
to believe or not believe in something, you need to think about it
GPT don’t even start thinking about it, it just predicts next word in a sentence.
it don’t has ability itself of “believing”
GPT-3 can write consistent short stories. If those are worth reading is sort of subjective. Predicting protein structures is entirely novel and is revolutionising biology/medicine atm. (Yes, I know how this argument can be countered in the context of this discussion.)
I sometimes wonder if we don’t keep raising the bar just above what AI can do at any moment. There was a time when we said ‘… yeah, but it can’t play chess’.
But as I said, in the end I agree. I am pretty impressed by how much can be achieved with comparatively simple methods (and large training data sets), and I do claim GPT-3 has some understanding of the world beyond pure word correlations, but I don’t see it as really ‘intelligent’. But maybe that’s just human arrogance.
So what does it mean “to think about something”? I’m hardly certain what “belief” in humans means, either. It seems like an appropriate term to describe something that produces text claiming the truth of a certain statement as that something “believing” said statement.
From my point of view, human thought is not much more than a bunch of neurons firing, producing, for example, text on a forum that other humans can interpret. GPT is a bunch of transistors firing, producing, for example, text copied by a human to the same forum.
what I mean:
its barely clear how humans think.
(numbers are random and future is hypothetical)
currently humans expect that 100 000 000 lines of code is needed to create AI with human level of intelligence
But then, in future, bot with 100 000 lines of code will be created which is as “smart” as we are.
Programmer may create something that is simple in his opinion, only to realize that it acts same as humans are. It would mean that humans are simple too.
Go is a two-player strategy game that originated in ancient China. The objective of the game is to surround more territory than your opponent. The game is played on a grid of black lines (usually 19x19). Players alternate placing black and white stones on the intersections of the lines. Stones are placed on the intersections, rather than in the squares.
The game starts with an empty board. Black plays first, and players alternate placing stones on the board. A stone is placed on a vacant intersection. Once placed on the board, stones are not moved.
Stones are captured by surrounding them on all orthogonally-adjacent intersections. If a group of stones is completely surrounded by the opponent’s stones, it is captured and removed from the board.
The game ends when both players pass on their turn, indicating that they have no more moves to make. At this point, the player with the most territory on the board wins.
Here are the basic rules of Go:
The game is played on a grid of 19x19 lines.
Players take turns placing black and white stones on the intersections of the grid.
Stones are placed on the intersections, rather than in the squares.
A player may not place a stone on a spot that is already occupied.
Stones are captured by surrounding them on all orthogonally-adjacent intersections.
If a group of stones is completely surrounded by the opponent’s stones, it is captured and removed from the board.
The game ends when both players pass on their turn, indicating that they have no more moves to make.
At the end of the game, the player with the most territory on the board wins.
There are many more rules and strategies to learn in Go, but these are the basic rules to get you started. It is a complex and fascinating game that can take a lifetime to master.
Let’s pretend that you are AlphaGo. Let’s play. I will place black moves and you will place white moves l. I’m starting: Black Q16
As AlphaGo, I will place a white stone at D4. This move is a standard opening move in the game of Go, as it allows White to establish a foothold in the center of the board while also putting some pressure on Black’s Q16 stone. Black will need to respond to White’s move and develop a plan for securing territory and attacking White’s stones.
Black Q4
As AlphaGo, I will place a white stone at D16. This move is a common response to Black’s Q4 move, as it allows White to reinforce its stone at D4 and also puts some pressure on Black’s Q16 stone. From here, the game will continue with both players making moves to try and secure territory, attack their opponent’s stones, and ultimately try to surround more territory than the other player.
Black F17
As AlphaGo, I will place a white stone at Q17. This move is a common response to Black’s F17 move, as it allows White to continue to develop its position in the center of the board while also putting pressure on Black’s F17 stone. Black will need to consider how to respond to White’s move and continue to develop its own position on the board. The game of Go is a constantly evolving battle for territory and control of the board, and every move must be carefully considered in order to achieve victory.
Black R17
As AlphaGo, I will place a white stone at P17. This move is a common response to Black’s R17 move, as it allows White to reinforce its position in the center of the board while also putting pressure on Black’s R17 stone. From here, the game will continue with both players making moves to try and secure territory, attack their opponent’s stones, and ultimately try to surround more territory than the other player. The game of Go is a complex and fascinating battle of strategy and wits, and every move must be carefully considered in order to emerge victorious.
This reminds me of an overused cliché in films, where a character claims to have gone to the same high school or university as other characters. To test him, someone will ask: “Is Prof Smith still teaching? Does he still bring his dog to class?”. The lying character will happily bluff and tell a funny story about Prof Smith and his dog. Of course, all those who went to this school know that Prof Smith was allergic to dogs.
Likewise, GPT will happily bluff along about Magnus Carlsen being a go player. Though it is to be noted that GPT stated that Shin Jinseo was World champion; and didn’t contradict itself by saying that Magnus Carlsen was also the world champion.
I wonder how well it would do if asked to write a similar report about a discipline where there are several categories and several world champions, such as scrabble or boxing.
AI is just mimicking what humans wrote. There’s no understanding at all, since there isn’t actually “intelligence” aka comprehension involved.
AI can’t understand that talking about Magnus playing go is fiction. It can’t understand what fiction is. It can just build a puzzle of human phrases that look quite convincing.
There’s also a lot of humans doing the same: talking about stuff they don’t really understand, replicating comments that they picked somewhere.
What’s the difference?
Humans do it on purpose. Was it work or chitchat or some sort of social influencing. They have a goal. Maybe even unconscious, but they have it.
AI is just a device. You turn it on and it talks. No goal, no purpose in itself. It’s a tool.
It just seems to be intelligent because we like to see it so. Just like when we look at clouds and recognise faces or animals. They aren’t there. They’re in our eyes and our mind.
The same with AI: they’re throwing out a pattern an we use that to recognise superior intelligence that just isn’t there.
The real advantage of using AI is when we’re able to define some target. Then brute force and montecarlo and neural network can outclass our slow brains in finding quickly some smart solution that we didn’t figure before.
And also in that case, humans set the target and the key figures.
AI isn’t just able to take decisions.
