Oh well, some concerns around AGI might be attempts at marketing and attracting attention, though some caution probably isn't too bad to have. AGI might become a thing, it's just that there's no guarantee that it will happen in our lifetimes. On the other hand, even LLMs will put some folks out of a job.
Relations between nuclear powers are anyone's guess but there's no reason why Russia couldn't be forced to back off, as opposed to "being backed into a corner". Wouldn't be the first proxy war and most likely won't be the last.
I care about climate change, there are certainly others. No idea whether much of a change will be made in our current course, but there's no reason to be super dismissive. If anything, recycling, eating a bit less meat or doing lots of the other common recommendations improves my own quality of life, even if the legislature that goes after the big corporations is nowhere to be seen yet. No reason not to make the world a better place, or ar least try to, in whatever ways are viable, like donating towards planting trees or something like Wren.
It is “just” a matter of combining. I think it’s an “engineering problem” by now. Which is not to say that it will happen soon, just that if we set our minds to it it won’t be that big of a deal.
The LLM's we have today weren't supposed to show up until 2030 at the earliest and 2050 more realistically. If you go back 10 years and read what people were writing about AI progress, you will find that they were _completely_ off the mark. It's prudent to assume a very liberal timeline for AI progress, based off whats been happening now.
Personally, since I first started to learn about Neural Nets in the 90's, I've always predicted that they would bring AGI around the time they matched human brains in complexity and compute power. Back then, assuming Moore's law meant that would be between 2030-40.
Now, there was always the prospect of a serious breakdown of Moore's Law, which might delay it, but since that didn't happen (compute power just moved from CPU's to GPU's), we're now approaching the final stretch. Human brain scale neural nets are already being planned, if not already being built in secret.
Not an achievement at all?
Talking to an ai?
Generating any image you like?
Etc.?
Please enlighten us sensei, what is actually General Intelligence, and how will you know it's here?
I realize my reply is a bit flippant, but it's tiring to see how certain are people about something that nobody truly understands.
AGI simply means that we can create a type of machine that can do any kind of mental task that a human can, at least at a human level.
> AGI simply means that we can create a type of machine that can do any kind of mental task that a human can, at least at a human level.
What does this even mean? What is human level? The ability to write a symphony? Or the ability to calculate 2+2=4? These are such poorly defined metrics that I don't understand how can we then start throwing around very precise (and short!) timelines for attaining it.
The statement: "we'll send a human mission to Mars in the next 5 years" is 10x more believable than this breathless AGI hype, because at least the former is well defined.
I would say that current AI's do have a kind of intelligence. It's just not generic enough that we label it AGI.
Add an ability to form, optimize an execute plans while also increasing the physical world model, and we're getting really close.
Both of these abilities are already being developed.
AGI doesn't require consciousness. We likely will have no idea whether or not the first AGI computer is conscious or not. What we will know though, is that it is solving novel problems that humans struggle with, and from the outside the only way to discern it isn't another human is because it is way to smart to be one.
I strongly suspect that the first computers which actually are conscious (AGI or not), whenever that happens, will have to fight a heavy uphill battle to prove it. And there will always be a group of people who will never believe an AI (is it even artificial at that point?) is conscious.
And I'm not particularly keen of the reasoning: "it is solving novel problems that humans struggle with", because a pocket calculator can solve problems that humans struggle with. A computer can play chess better than a human can, does it mean it's an AGI?
https://en.wikipedia.org/wiki/Artificial_general_intelligenc...
> failing to move away from the cost in the next 100 years
Unless you're a time-traveler from the 50s who has somehow managed to post here, there is no excuse for this type of disinformation these days.
Mentioning higher sea levels is also a red herring; massive agricultural yields collapse will be an issue long before (like, this century) sea levels become a major problem.
More than this, we have now reached levels of atmospheric changes that put actual near-term Human extinction (not to mention that of most sea and land species) on the table.
The IPCC reports say we might reach 4C, 5C or even more; based on the historical record, such a major change in such a short time - several orders of magnitude faster than previous CO2e-gases-linked mass extinction events - likely cannot be adapted to by the majority of species (there will of course also be a few evolutionary winners), resulting in potential extinction. I also quote the latest draft report from that same IPCC, leaked about two years back by concerned involved scientists to newspapers before the usual step where political stakeholders are allowed to reword the parts they deem too disturbing or against their interests:
"Life on Earth can recover from a drastic climate shift by evolving into new species and creating new ecosystems," it says. "Humans cannot."
Not to mention, at +5C it is all but certain shallow methane hydrate deposits (those stabilized by temperature, not pressure) all over the world are now outgassing CH4; not to mention several other similar tripwires, and likely not all of which we've even identified.
Edit: I can't seem to be able to reply to your comment below, not sure why. You're absolutely wrong about those Siberia figures, and they're not supported by current scientific consensus.
We're also talking about a time span of 100-300 years. Keep in mind how much more efficient farming is now compared to even the 1950's.
it only takes a few days (or hours) of above wet bulb temperatures to kill humans who live in and around the tropics today.
Still, this scenario is quite different from human extinction, which is what I was responding to further up.
> GHG emissions will lead to increasing global warming in the near term, and it’s likely this will reach 1.5°C between 2030 and 2035.
https://climatechampions.unfccc.int/the-ipcc-just-published-...
> Global temperatures have been exceptionally high over the past three months – at around 1.6C above pre-industrial levels – following the peak of current El Niño event at the start of 2024.
> The past 10 months have all set new all-time monthly temperature records, though the margin by which new records have been set has fallen from around 0.3C last year to 0.1C over the first three months of 2024.
https://www.carbonbrief.org/state-of-the-climate-2024-off-to...
2. Nuclear conflict is real, it's astounding Pakistan or Russia hasnt used the bomb yet but they will when the US backs them into a corner.
3. Climate change? At this point, who cares? We know it's happening and happening VERY slowly so we have PLENTY of time to get ready. Theres no real risk other than failing to move away from the cost in the next 100 years. International shipping and big ag are the largest polluters by far and those ain't stopping.