- cross-posted to:
- [email protected]
- cross-posted to:
- [email protected]
Another article for free (but seriously subscribe to 404 they do great work) https://futurism.com/artificial-intelligence/graduation-speaker-booed-ai
Another article for free (but seriously subscribe to 404 they do great work) https://futurism.com/artificial-intelligence/graduation-speaker-booed-ai
Radio to television is generous, IMO. LLMs aren’t a path to AGI or even particularly useful to most people.
I believe that’s the point for proponents of AI: compare it to something that was once considered amazing, hoping you’ll fail to realize that it’s like comparing apples to oranges.
Telegraph was the big one, that shortened broadcast communication times from weeks to a couple days. (Receiving the news telegram -> publishing and distributing the newspaper). News ticker, teletype, and eventually, telephone were all evolutionary ideas: they relied on the same newspaper for broad dissemination.
Radio was the next revolution, shortening news distribution from a couple days to a couple hours, bypassing the newspaper and going directly to the public. TV was a relatively small evolution of radio. It didn’t increase the speed or breadth of distribution; it only expanded the scope of what was distributed.
The internet was a big revolution. Cloud computing was another evolutionary idea. AI is a rather small evolutionary take off from that.
We are due for a course correction on Cloud Computing and Storage. Too many companies trust their most intimate secrets to Microsoft, Google, and Amazon; not to mention CloudFlare.
If I had to guess, too many government agencies are probably bought by big tech. Otherwise, they wouldn’t let that neglect of privacy fly at such a scale. I suppose the question is when people will get sick enough of it that it results in a change.
most of eu/usa is trying to give access to palantir to private info, Uk with NHS, and palantir trying to get SSN/flock cameras, or other.
The EU also seems to be trying to copy the UK Online Safety Act: https://gitlab.opencode.de/bmi/eudi-wallet/wallet-development-documentation-public/-/work_items/13
Once again, LLMs have to prove useful for something first. I wager they’re actually costing companies in the long run. Nothing about a net loss is revolutionary.
Any machine or algorithm that cannot consistently produce the same results from the same input is fucking garbage and that is LLM in a nutshell. Garbage.
A randomization function is expected to return a random result and works consistently.
Yeah, I was just playing around being annoyingly pedantic. I agree with your original statement, for what that’s worth.
AI (specifically the LLMs for which “hyper scale” datacenters are being built) are by good for summarizing large volumes of text down into a few points with reasonable accuracy.
This makes them good for pouring over all the data you’re being forced to store in the cloud, including your conversations, financials, political donation receipts, and that patent you’re preparing to file for your million-dollar idea that just might work. Please be sure to associate all your accounts with your verified ID or they will be subject to deletion and your name will be automatically listed on the suspected terrorist and pedophile watch lists.
Continue to participate in capitalism. We demand it.
And to have them be useful is a lot of setup and fuck around, needing to be done by someone experienced and not given to idiots to use for anything requiring reliability or rich substance.
I’d say they’re like going from 3.1 to 95, but web versions which most people use are pretty useless for non-basic things and basic things it’s usually faster to do yourself.
It’s just how algorithmic prediction is.