Quote (RedFromWinter @ May 9 2024 01:24am)

I work with AI tech, LLM, NN, LP techniques so I have enough understanding to be not be naively afraid or fear mongering with it.

Quote (thesnipa @ May 9 2024 07:20am)

is most of your exposure on the language side of automation? that stuff isn't as concerning to me in disruption yet. it will get there but its 20 years behind labor automation and more complex.

Quote (RedFromWinter @ May 9 2024 06:42am)

It's varied, mostly AI applications that are not driving physical devices, like stock exchange bot, huge network flow problem we solved for work. Have a background in EE/CE, but been doing enterprise software crap for a while. Easy money :)

The AI tech that blew my mind a few years ago was the facial recognition using neural networks and low resolution crappy cameras. A small network, low resolution input, and low end PC is enough for that to confidently match majority of faces up. Tensor flow has some free base models for this if you want to play.

As for physical automations that are truly independent like a humanoid bot, I think it's further away outside of niche use cases. If the battery tech was better it may come faster IMO.

Quote (thesnipa @ May 9 2024 08:50am)

i think a lot of tasks people mentally attribute to "needs a robot" are more solvable with a nonrobot than people think. i see it a lot in the packaging industry with my work, we solve some pretty crazy applications with just X/Y movements. and really its a macro calculation, 100 employees down to 10 is huge on a large scale of factories. we've done installs for top food manufacturers and we get personnel info, and you can just watch it dwindle project after project. we put in lines that do the exact same product and u can see the numbers sky dive.

Quote (RedFromWinter @ May 10 2024 03:18am)

This is a bit tangential to our back and forth, but one actual fear I have with technology and AI is an aspect of it is how humans are archiving our new information, data-sets, shared corpus. To a large degree,.modern information is locked in proprietary computer systems either by design or by technical constraints of the technology such as needing some goofy software to load and analyze said data. Obviously a programmer can feed this to an AI in batch. What about a human though without technical means? They would be lucky to find a academic journals hot take on said data in a library.

Point is, scientific information of the past was often written into journals you can actually check, scan, reverify. While painstakingly slow, paper and ink is a universal medium mostly standing test of time if copies allowed. In contrast, much of our recent corpus is in digital libraries. AI has huge advantage there while Human may be simply left out.

An easy example is modern Egyptian pyramid data and scans. Most of them unattainable. You could go copy some old French archeologist log books though from 1900s

Quote (RedFromWinter @ May 10 2024 05:18am)

This is a bit tangential to our back and forth, but one actual fear I have with technology and AI is an aspect of it is how humans are archiving our new information, data-sets, shared corpus. To a large degree,.modern information is locked in proprietary computer systems either by design or by technical constraints of the technology such as needing some goofy software to load and analyze said data. Obviously a programmer can feed this to an AI in batch. What about a human though without technical means? They would be lucky to find a academic journals hot take on said data in a library.

Point is, scientific information of the past was often written into journals you can actually check, scan, reverify. While painstakingly slow, paper and ink is a universal medium mostly standing test of time if copies allowed. In contrast, much of our recent corpus is in digital libraries. AI has huge advantage there while Human may be simply left out.

An easy example is modern Egyptian pyramid data and scans. Most of them unattainable. You could go copy some old French archeologist log books though from 1900s

Quote (ofthevoid @ May 10 2024 12:12pm)

Dystopian future in which all knowledge and history is digitized and can be manipulated in 'We were always at war with east Asia' Orwellian fashion. Or the alternative, even if it's not manipulated, it's vulnerable to complete knowledge loss if so much of it is stored on things that need electricity.

Quote (thesnipa @ May 10 2024 03:03pm)

imo its more likely we have too much access to info, not too little. we're already feeling the effects of that. whenever anything happens we see both sides of the aisle spin completely opposite narratives and many people eat it up.

Quote

Microsoft (MSFT.O), opens new tab and OpenAI are working on plans for a data center project that could cost as much as $100 billion and include an artificial intelligence supercomputer called "Stargate" set to launch in 2028, The Information reported on Friday.

OpenAI did not immediately respond to Reuters' requests for comment.

Rapid adoption of generative artificial intelligence technology has led to sky-rocketing demand for AI data centers capable of handling more advanced tasks than traditional data centers.
The Information reported that Microsoft would likely finance the project, which is expected to be 100 times more costly than some of the biggest existing data centers, citing people involved in private conversations about the proposal.

Quote (ofthevoid @ May 10 2024 01:19pm)

Yeah I see your point, trying to drink from a fire hydrant and still ending up thirsty.

One reason I think there could be some level of knowledge loss is at some point, there will be a certain aggregation and gatekeeping for some of these bigger AI platforms. For example Microsoft and OpenAI are saying that they will need several gigawatts of power to build and run one of these super-computers. It won't be hundreds of corpos doing this, but a few large behemoths that will monopolize data and at some point probably restrict it to some extent. Maybe they don't restrict it but simply leave the competition in the dust with no one using it, similarly to how Google did with Bing or other search engines during the last 20 years. So the information channels are basically squeezed into a few.




https://www.reuters.com/technology/microsoft-openai-planning-100-billion-data-center-project-information-reports-2024-03-29/

Quote (MildSambal @ May 6 2024 08:44pm)

I currently don't see a future where one of three things don't happen:

1) UBI is mandated.

2) Laws are passed that require companies to have X% of human workers.

3) A combination of the two.

UBI was a policy brought up in the 2020 election cycle in anticipation of LLM based technology (ChatGPT) and the possibility of AGI on the horizon

Where we are today, AGI is more in the realm of science fiction than reality but we aren't really sure how far off highly advanced models come to AGI.

AI will continue to clear out more and more menial tasks which will undoubtedly create higher rate of job displacement than new opportunities

By 2027 fully autonomous trucks, including truck platoons of two or more trucks in which all trucks have a driver, but only the driver of the lead truck has full control of the vehicle, are anticipated to appear on highways

Probably by the end of this decade a big majority of drivers will be automated

Other labor jobs such as construction workers, warehouse workers, cooks, janitors, mechanics, bartenders, electricians, plumbers, carpenters, service, etc will have a significant chunk automated

Even an unemployment rate of 10% would be significant enough to demand attention to policies such as UBI, and it is extremely difficult to imagine the U.S. not hitting at least 10% unemployment during the next administration