Best Of

@marcitko said:
Some musings from recent history...

Chess and poker and Go are games with a limited number of rules and clear objectives. For some time now, computers have been far better at these games than even the best humans. We say that these games have been "solved" (that the optimum strategy has been found) but this is actually not true - since computers keep improving and finding even better strategies.

What has happened with humans playing these games...
... the best players try to learn from and emulate the computer strategy. Generally, the closer you are to the computer strategy, the better you are. However, noone can fully do so because the computer strategy is too complicated.
... (I am not sure, but I strongly suspect) the average human skill-level in these games has increased. This is primarily because now the correct principles have been "proven" and so they can be learnt and taught in ever simpler and more efficient ways.
... some of the "mistique" has been lost for players at the top (because the computer is better).
... there is increasing concern about cheating (ie. using computers while playing other humans).
... people still want to see people play these games and play themselves. Pretty much noone watches computer vs computer games and human vs human games are as popular as ever.
... if your goal is to really delve deeply into the strategy aspect and learn, it is now easier than ever and you can go far further.

I too am concerned about AI, and am not sure what my comment adds to the discussion, but I guess computers solving games was a precursor to today's AI, so how that played out could be partially indicative of how AI will play out.

This all reminded me of a recent development that gives me a "Go Humans!" feel. Someone developed a simple strategy that anyone could easily beat the Go AI. Don't ask me to explain it, but basically its a strategy that any basic Go player would recognize and easily prevent if played against them, but the AI had no clue what was happening.

https://www.vice.com/en/article/v7v5xb/a-human-amateur-beat-a-top-go-playing-ai-using-a-simple-trick

What that kind of shows is that AI currently has no sense of what it is doing or what it is looking at, it is just a super sophisticated calculator that looks like it knows what its doing from the outside. Another way to say it is maybe AI is an example of hyper intelligence with zero wisdom or common sense.

Some musings from recent history...

Chess and poker and Go are games with a limited number of rules and clear objectives. For some time now, computers have been far better at these games than even the best humans. We say that these games have been "solved" (that the optimum strategy has been found) but this is actually not true - since computers keep improving and finding even better strategies.

What has happened with humans playing these games...
... the best players try to learn from and emulate the computer strategy. Generally, the closer you are to the computer strategy, the better you are. However, noone can fully do so because the computer strategy is too complicated.
... (I am not sure, but I strongly suspect) the average human skill-level in these games has increased. This is primarily because now the correct principles have been "proven" and so they can be learnt and taught in ever simpler and more efficient ways.
... some of the "mistique" has been lost for players at the top (because the computer is better).
... there is increasing concern about cheating (ie. using computers while playing other humans).
... people still want to see people play these games and play themselves. Pretty much noone watches computer vs computer games and human vs human games are as popular as ever.
... if your goal is to really delve deeply into the strategy aspect and learn, it is now easier than ever and you can go far further.

I too am concerned about AI, and am not sure what my comment adds to the discussion, but I guess computers solving games was a precursor to today's AI, so how that played out could be partially indicative of how AI will play out.

"We live in interesting times."
I would say we live in scary times. Look at how humans have handled the internet and social media ( hint: extremely badly). I don't see AI ending well for humans. Not trying to be negative but an honest take. Look at the rise of the far right across the globe. AI will undoubtedly also be put to terrible use by certain humans. But as we see in history it only takes a powerful minority to destroy the world. I am not an AI fan and do not embrace it. But totally can see why people might be more optimistic than I am about it.
Humans haven't learned the lessons of history which is why we are always on the brink of collapse. It's not the AI as such I don't trust but the group of humans who will undoubtedly use it for dangerous purposes.
I have many other reservations besides the one I've mentioned. I wish I could feel we were living in exciting times in a positive way but that's not how I feel.

Mainly I'm saying don't judge us all by the headline grabbing extremists.

It's true that there are wackjobs in most countries, it's just that in the US they seem to be more prevalent and due to weird and whacky worldwide news coverage from the US, the wackiness extremes come to the attention of the rest of the world more often that the wackjobs elsewhere...

I know that certain states have more wackjobs than others (Texas comes to mind)..The tight grip that warmongering apocalyptic evangelical Christians have on US politics is (for want of a better term) scary...

Quite a while back I was on a Buddhist forum (it's no longer running ) there was a Zen practitioner there who lived in an apartment block in a city in the bible belt, he never disclosed to his neighbours, friends and work colleagues that he was Buddhist, for fear he would be ostracised and seen as a devil worshipper, at first I thought he was joking, but he was serious...

There was an Nvidia conference on Wednesday, during which they announced a new chip code named Blackwell, which is the largest parallel processor they have built to date. When deployed in the data centers which power the cloud, it will enable the creation of AI models ten times as complex as the one in ChatGPT-3. It will take a couple of years for this to happen, but then we may find AI which is substantially more capable than human-level.

Nvidia’s technology is a kind of bedrock on which a lot of AI processing is built, and it is earning them a lot of money. They used to build only graphics processors, which earnt them 2.9 bn USD a quarter, but that is now dwarfed by the business for AI processors, which earnt them 18.7 bn USD in the last quarter.

Can't say when AGI will come but since the quick rise of self driving cars then a stall, I've heard what sounds like a fairly common coding saying, "99% of the work is the last 1%".

So I guess my uninformed opinion would be that these sorts of specific AIs will proliferate and get better, but AGI will be a ways off yet.