And Stargate Project is... what exactly? What is the goal? To make Altman richer, or is there any more or less concrete goal to achieve?
Also, few items for comparison, that I googled while thinking about it:
- Yucca Mountain Nuclear Waste Repository: $96B
- ITER: $65B
- Hubble Space Telescope: $16B
- JWST: $11B
- LHC: $10B
Sources:
https://jameswebbtracker.com/jwst/budget
Be the definitive first past the post in the budding "AI" industry.
Why? He who wins first writes the rules.
For an obvious example: The aviation industry uses feets and knots instead of metres because the US invented and commercialized aviation.
Another obvious example: Computers all speak ASCII (read: English) and even Unicode is based on ASCII because the US and UK commercialized computers.
If you want to write the rules you must win first, it is an absolute requirement. Runner-ups and below only get to obey the rules.
It has been quite clear for a while we'll shoot past human-level intelligence since we learned how to do test-time compute effectively with RL on LMMs (Large Multimodal Models).
Look, making up a three-letter acronym doesn't make whatever it stands for a real thing. Not even real in a sense "it exists", but real in a sense "it is meaningful". And assigning that acronym to a project doesn't make up a goal.
I'm not claiming that AGI, ASI, AXY or whatever is "impossible" or something. I claim that no one who uses these words has any fucking clue what they mean. A "bomb" is some stuff that explodes. A "road" is some flat enough surface to drive on. But "superintelligence"? There's no good enough definition of "intelligence", let alone "artifical superintelligence". I unironically always thought a calculator is intelligent in a sense, and if it is, then it's also unironically superintelligent, because I cannot multiply 20-digit numbers in my mind. Well, it wasn't exactly "general", but so aren't humans, and it's an outdated acronym anyway.
So it's fun and all when people are "just talking", because making up bullshit is a natural human activity and somebody's profession. But when we are talking about the goal of a project, it implies something specific, measurable… you know, that SMART acronym (since everybody loves acronyms so much).
Also, "Dario Amodei says what he has seen inside Anthropic in the past few months leads him to believe that in the next 2 or 3 years we will see AI systems that are better than almost all humans at almost all tasks"
“Preprint out today that tests o1-preview's medical reasoning experiments against a baseline of 100s of clinicians.
In this case the title says it all:
Superhuman performance of a large language model on the reasoning tasks of a physician
Link: https://arxiv.org/abs/2412.10849”. — Adam Rodman, a co-author of the paper https://x.com/AdamRodmanMD/status/186902305691786464
—-
Have you tried using o1 with a variety of problems?
Am I to conclude that we've had a comparably intelligent machine since 2012?
Given the similar performance between GPT4 and O1 on this task, I wonder if GPT3.5 is significantly better than a human, too.
Sorry if my thoughts are a bit scattered, but it feels like that benchmark shows how good statistical methods are in general, not that LLMs are better reasoners.
You've probably read and understood more than me, so I'm happy for you to clarify.
Perhaps it’s better that you ask a statistician you trust.
If not, why is the study sufficient evidence for the LLM, but not sufficient evidence for the previous system?
Again, it feels like statistical methods are winning out in general.
> Perhaps it’s better that you ask a statistician you trust
Maybe we can shortcut this conversation by each of us simply consulting O1 :^)
From a national security PoV, surpassing other countries’ work in the field is paramount to maintaining US hegemony.
We know China performs a ton of corporate espionage, and likely research in this field is being copied, then extended, in other parts of the world. China has been more intentional in putting money towards AI over the last 4 years.
We had the chips act, which is tangentially related, but nothing as complete as this. For i think a couple years, the climate impact of data centers caused active political slowdown from the previous administration.
Part of this is selling the project politically, so my belief is much of the talk of AGI and super intelligence is more marketing speak aimed at a general audience vs a niche tech community.
I’d be willing to predict that we’ll get some ancillary benefits to this level of investment. Maybe more efficient power generation? Cheaper electricity via more investment in nuclear power? Just spitballing, but this is an incredible amount of money, with $100 billion “instantly” deployed.
We're not spending money on AI as a field, we're spending a lot of money on one, quite possibly doomed, approach.
i guess what i'm asking is: what was the practical advantage of ascii or feet and knots that made them so important?
An almost absolute incumbency advantage.
>what was the practical advantage of ascii or feet and knots
Familiarity. Americans and Britons speak English, and they wrote the rules in English. Everyone else after the fact needs to read English or GTFO.
Alternatively, think of it like this: Nvidia was the first to commercialize "AI" with CUDA. Now everyone in "AI" must speak CUDA or be irrelevant.
He who wins first writes the rules, runner-ups and below obey the rules.
This is why America and China are fiercely competing to be the first past the post so one of them will write the rules. This is why Japan and Europe insist they will write the rules, nevermind the fact they aren't even in the race (read: they won't write the rules).
2) As mentioned in the chart label, earlier systems require manual symptom extraction.
3) An important point well articulated by a cancer genomics faculty member at Harvard:
“….Now, back to today: The newest generation of generative deep learning models (genAI) is different.
For cancer data, the reason these models hold so much potential is exactly the reason why they were not preferred in the first place: they make almost no explicit data assumptions.
These models are excellent at learning whatever implicit distribution from the data they are trained on
Such distributions don’t need to be explainable. Nor do they even need to be specified
When presented with tons of data, these models can just learn, internalize & understand…..”
More here: https://x.com/simocristea/status/1881927022852870372?s=61&t=...
A vessel traveling at 1 knot along a meridian travels one minute of geographic latitude per hour.
on the previous examples i can see language gave native speakers and advantage in becoming familiar with the technology but on ai i'm not seeing an advantage that would give americans an advantage over everyone else, besides controlling access to the tech.
the reason i'm insisting on this is because i feel as if that argument has merit but i have yet to grasp how it applies to these technologies.
The microprocessors concerned are very high value goods, manufacturing and R&D for them can't be easily and quickly spun up on a whim. The country and companies first to start them up and win will secure the supply chains, and once secured it will take monumental money and effort to reconfigure them. A lot of money is at stake, in other words.
Geopolitically, it also means that the country who secures the supply chain also gets to quite literally write the rules regarding who and where the microprocessors can be sold to and exported. Either the US or China gets to decide who can buy the microprocessors depending on who wins the supply chain.
Just like Nvidia was the first past the post and now enjoys absolute incumbency advantage, whichever country (namely US or China) is first past the post in the "AI" industry will enjoy absolute incumbency advantage.
i have to say the ascii, feet, and knots were a bit confusing though. these do not seem to be the same kinds of "wins" as what we're expecting to see with this race though. utf8 is mostly the default around the world and airbus is a serious competitor in international markets.
America to this very day gets to dictate how computing and aviation work. Knots, feet, ASCII and so on are just the obvious signs of that.
>utf8
Case in point, UTF-8 (aka Unicode) has ASCII as its starting point. ASCII can be converted to UTF-8 without data loss easily and perfectly because the first entries in Unicode are literally ASCII mappings. This is the virtue of winning first and getting to write the rules.
>airbus is a serious competitor in international markets.
And yet everyone outside of China and Russia still fly using knots and feet, that includes Airbus.