Not gonna happen, what everyone is calling "AI" isn't even actually AI, its just a type of predictive text
Technology
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
A lot of work is already just generating boring text and images
Not to mention the fact that a lot of people don't function on a deeper level than that in they daily lives
Did you pay attention to human “intelligence” lately? I think we are already there.
Well yes, AI as we know it. What he's saying that there's reason to believe certain advancements can be made that would put AI in a place capable of competing with human intelligence.
AI as we know it now is not what he's referring to.
Edit: Also, logically, if we keep things centered on strict circumstances (like performing repetitive tasks) then there are absolutely people who exist that an AI could outperform. That is a simple reality of having limiting disabilities.
To create AI, you first need to find a way for them to communicate with the user.
LLMs are just that, the first step.
The rest will follow. Maybe not in 5 years, but definitely in the near future.
Reads just like crypto hype. “The killer application is right around the corner! It’s almost here, I swear!”
You can't deny the amount of crazy progress in machine learning in the past 3 years. The hype is pretty understandable, far more than crypto ever was.
And like crypto it eats GPU and electricity more than people seem to understand. It can work 24/7 though, when/if it arrives.
The only thing Jensen sees is him milking the market.
“AI” is just a complex language model. It understands absolutely nothing. He is dead wrong and there is plenty of proof out there to prove him wrong.
He is the manufacturer of the hardware LLMs are trained on, he’ll say anything that will increase sales.
That's what I was thinking. I foresee him retiring on fat stacks of cash in 5 years, that he made by pumping up the stock price with claims he won't have to worry about fulfilling
"Don't be a prospector, be the person selling goods to prospectors."
~ paraphrase of an idea we've all likely heard
Although you are correct in your assertions about AI you forget how stupid people are. He may turn out to be right but not for the reasons he thinks he is.
Right, he isn't saying it to us, but folks who would invest in the company due to their majority position in the processor space powering the current LLM's.
Ah, of course. Well done.
He could be dead wrong but not because some internet comment says so. Maybe cite the proof you speak of…
Maybe Huang should.
I have been saying, that automation should be taxed for years now, and people hate it. The poem "First they came ..." comes to mind.
Taxed? Fuck that. Any AI with human intelligence deserves at least minimum wage.
But automation encompasses more than just AI with human intelligence. For the other cases, it should be taxed and the money used to fund more social nets.
Yes all the big tech companies are in desperate need of more money, the poor things.
How would it get measured?
Would should automation be disincentivized?
Just tax revenue and wealth.
I spent a little while thinking about this earlier in the year, I had the idea formed more cogently at time but I'll try and put it as best as I remember. Income tax can kind of be restated as a tax on a corporation as a function of the value an individual provides to the company.
This isn't perfect but, I'm a PAYE employee, so the income tax I pay is done so at source. I don't ever see that money. In real terms it makes little difference to me whether I pay zero income tax and the company reduces my salary but pays a fee to the government for the privilege of employing me. The tax rates don't change hugely over time and I'm not on the margins of a tax band, so this mostly holds true for me. My salary and the tax band that puts me in are a proxy for the value I provide to the company (under the assumption I make net positive money for the company).
I feel like an explicit change to codify this is required to allow for the proper taxation of companies undergoing a shift to automation, otherwise it's too easy to domicile profits/wealth elsewhere (as it stands). Even thinking about this now, for knowledge work, how do you tax a company in Germany when the processing is happening on a privacy compliant server in Somalia? Even more stringent data protection and localisation laws? Can your models cross borders? Does that lead to multi-tier AI based on the capabilities of underlying populations and availability of training data?
Generally I'm pro-humans not having to grind to live and I generally see AI/automation as a boon for this - alongside proper taxation and redistribution of wealth, but I'm not sure I've ever seen any good explanation of how the nitty gritty of this functions in the real world.
Look around the world. In poor countries, productivity is low. There are not many machines. People do a lot of manual labor. Rich countries have lots of automation.
If you want to live in a country with less automation, moving is an option. Migrating from a rich to a poor country is much easier than vice versa. But if that looks unappealing, then taxing automation should also be unappealing.
Working less isn't horrible. The OECD estimates that an average employee in the USA works 1811 hours per year. In Germany, it is only 1341, You can always volunteer in a non-profit if you feel you don't have enough to do. There's nothing to be afraid of. I don't even know why or on what Americans work so much. It feels like they spend half the office day on social media, complaining that they can't afford things.
Because knowledge work is never ending. There's always more to do.
As an American, I've worked with lots of my European counterparts over the years, and trying to get things done can be downright painful.
We're across multiple time zones, and Europeans refuse to be on a call that isn't in their typical work hours.
Kind of problematic when there's no one "time" where a Central Time American can be on a call during his work hours while a Brit, German, and an Estonian do to.
Multiple people will have to be flexible here, and assuredly it won't be our Western Europe peers.
There are things like change windows, to reduce risk of downtime for users. Those are established by when the users utilize the resources being changed. Sometimes that means I work a normal day, and get back on things at midnight or 2am to make a change and validate it. It needs to be done then, it's important, it's been entered into a massive scheduling system which tracks resources: subcontractor time, staff time, access to things like VM hosts to ensure our change doesn't conflict with other changes to shared hosts/network/power, etc. Many internal and external organizations can be involved in changes, the external generally incur additional cost, so we try to combine as many changes as possible to minimize that cost.
This is just one small example of the coordination involved in herding the cats of large infrastructure.
SMB is much easier, far fewer people and system impacts, practically no change management, so if something happens days later, tracing it back to those changes can be difficult or impossible. It's more wild-west, with knowledge retained in a small set of admins. Even there it can take many conversations between local power, remote power, subcontractors, vendors, telco, cloud providers, etc to manage changes. These can all be geographically disparate (I have a friend with a client with operations in CA, CO, NM, WV, MO). That's 3 time zones, with vendors, subcontractors, and contracts in all of them, under varying legal jurisdictions and regulatory domains. Something as simple as updating/replacing a remote monitor cell router can take months of conversations. Without the upgrade, they're in violation of state and federal regulations, with fines that can be $10k/day or more.
Just because you have no idea what other people do, doesn't make it any less important or valuable. Any boss is very appreciative when you stay on a call "past 5" to help prevent being fined like that. (I've been on calls that lasted 24hrs+, over Thanksgiving).
I just received this 6 day old post as new. I guess that's due to the issues with federation.
I'm not really sure what you are going for here. Are you saying that Americans need to work more hours to make up for the slack of Europeans?
Automation wont stop because of taxes... There needs to be money, for the people that loses jobs to automation. The products wont get cheaper with more automation.
I wouldn't want to move to a third world country like america, where the low taxes that are paid by the little guy, are used to help the big guy. I'm fine living in a country, where my relatively high taxes can make the country even better.
Imagine you want to produce something. Maybe you want to bake a couple 10,000 breads over the next few years. Whatever. You could hire 50 guys and buy some simple tools, or you could hire 5 guys and buy some advanced machinery. What do you do?
The typical business will pick the cheaper option. It will replace as much labor with machines as is cost-effective. A few businesses will make a thing out of being inefficient and expensive, like how Rolls-Royce cars are handmade.
If you tax automation, you make the machines more expensive. So, when someone has the choice between using machines or using labor, then it will be labor more often. So, you're right: It won't stop automation. You will just have less of it. Productivity will be lower. The country will be always be poorer than without such a tax.
People in such a country will either have to work more hours for the extra labor needed or do with less.
This article resembles this one beat for beat from a month earlier: https://www.cnbc.com/2023/11/29/nvidia-ceo-ai-will-be-fairly-competitive-with-humans-in-5-years.html
Is that because AI is getting smarter, or because humans are getting dumber?
Yes
"I have created a pretrained transformer LLM statistically indistinguishable from your grandpa posting on Facebook"
No
AI doesn't have to compete. It just has to do what computers do best: repetitive tasks.
Good, can we have one that’s 100 times smarter and just runs the government benevolently? I’m really sick of civic responsibility
Gee I wonder why the bloke with a very vested interest in seeing AI take off would make claims that AI will be super duper good you guys in the near future.
CEO of company creating AI overhypes AI. Film at eleven.
In Jensen's wet dreams