ChatGPT AI again, starting with 4 7 swap.

[BOOT]

2017

https://www.independent.co.uk/life-styl ... 69706.html

[BOOT]

Oh and what if you helped a party get elected via electronic voting machines...

[BOOT]

"Warning of AI’s danger, pioneer Geoffrey Hinton quits Google to speak freely"

https://arstechnica.com/information-tec ... ak-freely/

[NewbVetteGuy]

it can be a useful modification for certain applications. However, it's important to consider all aspects of the engine design and performance characteristics before making any modifications to ensure that the benefits outweigh any potential drawbacks.

You can definitely tell that it was fed a lot of legal training data and that it was trained early to be a "lawyer"...

Ask it if it knows that you're not paying it by the word / hour...


Adam

[NewbVetteGuy]

Weird Elon wants to ban this stuff.

It's because he subscribes to the Ray Kurzweil AI philosophical "School", which believes that the current "Single Purpose" AIs, evolve eventually to become "General Purpose AI" and that they just continue to advance exponentially until a magic "Singularity" event in which "Machine Intelligence" and "Human Intelligence" "merge" and then x seconds or minutes later, because of the exponential growth of knowledge of AI, the AI has all knowledge of the universe and instantly becomes an all-powerful god capable of literally altering time, space, and all matter. -Seriously and literally...

I had a co-worker a few years ago who went to Ray Kurzweil's "Singularity University" and was into all this stuff and understood it, but wasn't "SOLD" on where Ray ultimately ends up.


The fact that most of these same people seem "positive enough" that this all-powerful AI god would probably-hopefully view humans as worth keeping alive; well, at least sure enough to keep working towards making the all-powerful AI god happen, is pretty terrifying, too.



At least Elon is like "hold on; I believe all this AI god mumbo-jumbo, I just don't believe it's not going to kill us all. I'd probably kill us all, if I was it..."

Adam
P.S. Ray Kurzweil estimates the Singularty will happen around 2045, so we all have 23 years left to live, at most if he's right; lol!

[NewbVetteGuy]

Imagine an AI much more powerful in processing speed and you have something God level.

I'm not convinced, BUT.

IMHO, one of the best sources of information on where AI is going and how it will impact many of the emerging sectors of the economy and the exponential processing power that will be available to these thigns and the exponentially decreasing costs and time required to train them regularly come from ARK Investing's "BIG Ideas" report each year.

Definitely worth a read, even though it's nearly impossible to agree with all of their conclusions, there's some good research on trends:
https://research.ark-invest.com/hubfs/1 ... _Final.pdf

Some big highlights:

• AI is at the center of all emerging technology areas that are likely to experience exponential growth and will accelerate the other technologies (Robotics, Energy Storage, Multiomic Sequencing, Public Block Chains (Yes ARK is obsessed with Bitcoin)); ultimately they believe that cloud tech, self-driving cars, battery improvements, bio engineering, and the future of manufacturing will be driven by AI advances.
• The economic potential value of AI is greater than the economic potential value of the Internet
• The cost to train a large AI model, and for the hardware to run it will decline at a 70% annualized rate (in 2020 it cost $4.6 million to train a Chat GPT-3 level model, in 2022 only $450k, by 2030, it will cost only $30. (Cententralized and decentralized dedicated AI acceleration chips help vastly)
• Scary: The primary expected bottleneck to AI growth is access to enough training data. Ark estimates that the largest tech companies have access to the most training data, which will make their models superior, which will attract more customers, which will make their AI better still. -Big tech "eats" much of the global economy and it's basically impossible for any new-comers to ever compete meaningfully. (current trend accelerates: Think 50 or 100 Trillion dollar big tech company valuations vs. 2 Trillion...)
• Scary Adam Insertion: Eventually one of the "AI Ethics 101" rules of "Never Train your AI / Neural Network live connected to the internet" will almost CERTAINLY be broken to get around the "access to training data" issue AND so that the AI models can actually respond to "customers" inquiry's with up-to-date information. (CHAT GPT is trained I think ever 6 months with tons of data downloaded FROM the internet, which means at most it's always 6 months "behind".) -I mean I hope people just supplement the built-models with real-time services that can "fake" that the AI knows current information, but SOMEONE is going to connect one to the internet, if they haven't already.
• This one for Boot to make him even more paranoid: AI leads to a 10-fold increase in coding productivity (with AI essentially doing most of the "Code Authoring" in the world by 2030...) -Now it's authoring "The Matrix"! lol!

Agree or not, it's a heck of a read!

Adam

[NewbVetteGuy]

Here's a fun one for this crowd (attached).

Cost of transportation.png

Adam