Future of Artificial Intelligence

The "other" ERE. Societal aspects of the ERE philosophy. Emergent change-making, scale-effects,...
bostonimproper
Posts: 574
Joined: Sun Jul 01, 2018 11:45 am

Re: Future of Artificial Intelligence

Post by bostonimproper »

Apparently Bard is using Gemini as of yesterday. Played around with it a bit, I continue to be pretty unimpressed.

daylen
Posts: 2466
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Future of Artificial Intelligence

Post by daylen »

It's the pro version (as opposed to ultra) and doesn't include real-time multi-modal yet like in the video. Though, clearly if something like the demo can be mass distributed then education will undergo a paradigm-shift.

User avatar
fiby41
Posts: 1609
Joined: Tue Jan 13, 2015 8:09 am
Location: India
Contact:

Re: Future of Artificial Intelligence

Post by fiby41 »

TechCrunch| Google’s best Gemini demo was faked https://techcrunch.com/2023/12/07/googl ... -was-faked

Here are some more relevant reading materials if one wants to take a deep dive:

Gemini: A Family of Highly Capable Multimodal Models
https://storage.googleapis.com/deepmind ... report.pdf
AlphaCode 2 Technical Report
https://storage.googleapis.com/deepmind ... Report.pdf
Evaluating Large Language Models Trained on Code
https://arxiv.org/pdf/2107.03374.pdf
Google Secure AI Framework Summary
https://services.google.com/fh/files/bl ... ummary.pdf
MobileNetV2: Inverted Residuals and Linear Bottlenecks
https://arxiv.org/pdf/1801.04381.pdf
REAL TOXICITY PROMPTS: Evaluating Neural Toxic Degeneration in Language Models
https://www.semanticscholar.org/reader/ ... 17e876d21f

daylen
Posts: 2466
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Future of Artificial Intelligence

Post by daylen »

It's not that surprising given the pressure on google to respond. We should take any demo by these large companies to be partially staged. Though, do we really think google or these other players aren't capable of engineering something like this in the near future? Like one or two years max? Sure there will be many hurdles along the way, but the competition is out of control at this point.

daylen
Posts: 2466
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Future of Artificial Intelligence

Post by daylen »

Thousands of AI Authors on the Future of AI:
https://arxiv.org/abs/2401.02843
In the largest survey of its kind, 2,778 researchers who had published in top-tier artificial intelligence (AI) venues gave predictions on the pace of AI progress and the nature and impacts of advanced AI systems The aggregate forecasts give at least a 50% chance of AI systems achieving several milestones by 2028, including autonomously constructing a payment processing site from scratch, creating a song indistinguishable from a new song by a popular musician, and autonomously downloading and fine-tuning a large language model. If science continues undisrupted, the chance of unaided machines outperforming humans in every possible task was estimated at 10% by 2027, and 50% by 2047. The latter estimate is 13 years earlier than that reached in a similar survey we conducted only one year earlier [Grace et al., 2022]. However, the chance of all human occupations becoming fully automatable was forecast to reach 10% by 2037, and 50% as late as 2116 (compared to 2164 in the 2022 survey).
Most respondents expressed substantial uncertainty about the long-term value of AI progress: While 68.3% thought good outcomes from superhuman AI are more likely than bad, of these net optimists 48% gave at least a 5% chance of extremely bad outcomes such as human extinction, and 59% of net pessimists gave 5% or more to extremely good outcomes. Between 38% and 51% of respondents gave at least a 10% chance to advanced AI leading to outcomes as bad as human extinction. More than half suggested that "substantial" or "extreme" concern is warranted about six different AI-related scenarios, including misinformation, authoritarian control, and inequality. There was disagreement about whether faster or slower AI progress would be better for the future of humanity. However, there was broad agreement that research aimed at minimizing potential risks from AI systems ought to be prioritized more.
There is essentially a race going on to alleviate demographic collapse in many parts of the developing world. Globalization is reversing and this could lead to cascading disruptions in semiconductor production and renewables. How long before tesla or one of their new competitors starts leasing robots to work in the upcoming U.S. factory and mining boom? Necessity gives way to innovation.

jacob
Site Admin
Posts: 15767
Joined: Fri Jun 28, 2013 8:38 pm
Location: USA, Zone 5b, Koppen Dfa, Elev. 620ft, Walkscore 77
Contact:

Re: Future of Artificial Intelligence

Post by jacob »

daylen wrote:
Fri Jan 12, 2024 3:04 pm
There is essentially a race going on to alleviate demographic collapse in many parts of the developing world. Globalization is reversing and this could lead to cascading disruptions in semiconductor production and renewables. How long before tesla or one of their new competitors starts leasing robots to work in the upcoming U.S. factory and mining boom? Necessity gives way to innovation.
Loosely speaking, about 1/3 of the revenue of a company goes back into the company (to pay for goods, offices,...), 1/3 to labor, and 1/3 to investors. Insofar AI replaces 10% of the labor force, we're only talking increasing the margin by ~3% (10% of 1/3). Unless AI is going to open up an entirely new industry at scale that driving, plumbing, and electricity did---and which personal computing and the internet has only so-so done, I'm not going to be holding my breath that this will radically change the world at the Thomas Edison level.

zbigi
Posts: 933
Joined: Fri Oct 30, 2020 2:04 pm

Re: Future of Artificial Intelligence

Post by zbigi »

daylen wrote:
Fri Jan 12, 2024 3:04 pm

There is essentially a race going on to alleviate demographic collapse in many parts of the developing world. Globalization is reversing and this could lead to cascading disruptions in semiconductor production and renewables. How long before tesla or one of their new competitors starts leasing robots to work in the upcoming U.S. factory and mining boom? Necessity gives way to innovation.
There's plenty of robots in factories already. Autoplants make entire bar body (that includes pressing, welding, painting) with barely any human involved - it's done by a series of robotic arms. The humans are there mostly to supervise the robots and to do some periodic maintenance and cleaning.
Car factories also have simple self-driving cars operating on premises - they drive around and distribute parts to work stations.

7Wannabe5
Posts: 9261
Joined: Fri Oct 18, 2013 9:03 am

Re: Future of Artificial Intelligence

Post by 7Wannabe5 »

One of the books I am currently reading, "The Growth Spiral: Money, Energy, and Imagination in the Dynamics ofthe Market Process" by Binswanger, emphasizes that the primary purpose of capital is the promotion of the division of labor. So, it might be interesting to consider what may happen to the concept of "the firm" as AI becomes more generalist.

My 83 year old semi-disabled (can hobble her way to the bathroom) mother pays somebody to help her with cleaning 4 hours per week, but a "general purpose" human could help her with much more complex tasks with the assistance of an AI with generalized intelligence applied to a variety of skill sets. For instance, an AI, or a very basic AI home robot*, working in conjunction with a "general purpose" human (perhaps only needed to help dig out some paperwork to be scanned by AI)could prepare my mother's tax return. It could also provide the "general purpose" human with step by step instructions for fulfilling any whim of my mother, such as "Make me a Paris Hilton Spa diet lunch." or more critical tasks, such as "hook up electrodes for heart health analysis."

At one level up, a generalized AI could also greatly expand the functionality of a small machine shop or maker space, because it could provide the "brain" aspect of any brain/body skill set. IOW, best case scenario, it could provide much of the efficiency gains provided by high IQ and mental experience to a much larger proportion of the population. Therefore, it could theoretically make ERE (or similar)more accessible to more of the population.

*https://www.youtube.com/watch?v=MfLdhzc2oWw

The way this robot keeps an eye on pets could, obviously, also work for disabled or confused seniors, who will become more abundant with the demographic shift. Samsung hasn't released pricing yet, but even at cost of several thousand dollars it would be worthwhile purchase just for performing this task. For instance, on more than one occasion my mother has called fire department to pick her up off the floor, and the extent to which such an incident was a true emergency could be better ascertained with the robot's ability to send live video feed to emergency services and family members.

jacob
Site Admin
Posts: 15767
Joined: Fri Jun 28, 2013 8:38 pm
Location: USA, Zone 5b, Koppen Dfa, Elev. 620ft, Walkscore 77
Contact:

Re: Future of Artificial Intelligence

Post by jacob »

7Wannabe5 wrote:
Sat Jan 13, 2024 9:30 am
One of the books I am currently reading, "The Growth Spiral: Money, Energy, and Imagination in the Dynamics ofthe Market Process" by Binswanger, emphasizes that the primary purpose of capital is the promotion of the division of labor. So, it might be interesting to consider what may happen to the concept of "the firm" as AI becomes more generalist.
I forget where and what, but I read a rather long article written by a senior programmer (specialist) comparing his skill set to an "enthusiastic" hacker (generalist). It was his opinion that AI (in its current form) benefited the hacker more than the generalist in that AI could fill out gaps in the hacker knowledge. The example given was some hardware project wired into an iPhone app. Apparently learning the iPhone eco-system requires too much mental investment to just hack your way through it. However, with AI assistance the hacker friend was able to build a functional prototype app within one day.

In my experience, the usual problem of a generalist is really some kind of skill (or part) bottleneck---some small issue---that must be resolved in order to continue. Insofar AI can remove such "hard won training" bottlenecks, it could hurt a lot of specialists while benefiting a lot of generalists. Compare the two figures in chapter 4 of the ERE book.

7Wannabe5
Posts: 9261
Joined: Fri Oct 18, 2013 9:03 am

Re: Future of Artificial Intelligence

Post by 7Wannabe5 »

@jacob:

I absolutely agree. Inspired by my own post above, I prompted AI with the following:
How could I make a dress like Paris Hilton might wear using only materials I might find in a dumpster?
and the response was the best I've ever generated:
Creating a dress like Paris Hilton might wear using materials from a dumpster is a creative challenge! Let’s channel our inner fashionista and imagine a dumpster-diving couture masterpiece:

The Dumpster Chic Gown:

Materials Needed:
Discarded fabrics (old curtains, bed sheets, or tablecloths)
Broken jewelry or metallic scraps
Abandoned lace or tulle
Twine or shoelaces
Imagination (the most essential material!)

Construction Steps:
Base Fabric: Find a large piece of fabric (maybe an old curtain) as the base. Drape it around your body to create a loose silhouette.
Asymmetry: Tear or cut the fabric unevenly to achieve that avant-garde look. Remember, perfection is overrated!
Accessorize: Scour the dumpster for broken necklaces, earrings, or metallic scraps. Attach them to the dress for a touch of bling.
Lace Details: If you find discarded lace or tulle, layer it over the base fabric. Imperfections are welcome!
Belt It: Use twine or shoelaces as a makeshift belt. Knot it loosely around your waist.
Hem Drama: Let the fabric trail on the ground. Extra points if it collects a bit of dumpster dust!
Confidence: Strut down the imaginary runway with confidence. Remember, fashion is about attitude.

The Dumpster Diva Accessories:

Materials Needed:
Bent wire hangers
Scraps of fabric or plastic bags
Bottle caps
Broken sunglasses
More imagination (seriously, it’s the key!)

Construction Steps:
Wire Crown: Bend wire hangers into a crown shape. Wrap fabric scraps or plastic bags around it for a regal touch.
Bottle Cap Earrings: String together bottle caps (cleaned, of course) using twine or thread. Voilà, statement earrings!
Sunglasses Necklace: Remove the lenses from broken sunglasses. Attach them to a shoelace or twine. Wear it as a quirky necklace.
Mismatched Bracelets: Stack mismatched bracelets made from shoelaces, fabric strips, and anything else you find.
Own It: Stride through the dumpster alley like you’re on a red carpet. Paparazzi? Nah, just raccoons.

Remember, fashion is about self-expression, and sometimes the most unconventional materials lead to the most memorable looks. Paris Hilton would be proud of your dumpster-diving style! 🌟🗑️
The bit about using the lenses from broken sunglasses to substitute for jewels is fantastic. Although, I would have to follow up with technical question about how to create holes in lenses for stringing. I also appreciated the humorous substitution of raccoons for paparazzi. I used the moderately creative setting on the AI for this prompt. A robot or device with camera for image input could greatly extend the advice offered as I created or modeled the dress. Meanwhile, my solar powered garden robot working in conjunction with my geo-thermal powered artificial cow might be out foraging for high protein greens to be converted to tofu like substance for my dinner.

7Wannabe5
Posts: 9261
Joined: Fri Oct 18, 2013 9:03 am

Re: Future of Artificial Intelligence

Post by 7Wannabe5 »

Actually, "general purpose" human not even needed. New housekeeping robot via Stanford and Google. Estimated price less than $10,000 in mass production:

https://www.youtube.com/watch?v=HaaZ8ss-HP4

The human era is over.

zbigi
Posts: 933
Joined: Fri Oct 30, 2020 2:04 pm

Re: Future of Artificial Intelligence

Post by zbigi »

7Wannabe5 wrote:
Sat Jan 13, 2024 1:50 pm

The human era is over.
Similar things were said about self-driving cars less than 10 years ago... Just like with self-driving cars, having actually done some work in robotics, I'll believe it when I see it.

Specifically, the video you included is not even driven by AI (the description says it's tele-operated). The link to an actual AI-driven robot that they give (https://www.youtube.com/watch?v=HaaZ8ss-HP4) says that it employs "behavior cloning", which sounds like it's basically copying the movements performed earlier by the operator (the training set). How this translates to environment in any way different from the training lab they used is an open question.

User avatar
Ego
Posts: 6326
Joined: Wed Nov 23, 2011 12:42 am

Re: Future of Artificial Intelligence

Post by Ego »

The future is here.

Original dubbed speech by Javier Melei at Davos yesterday: https://www.youtube.com/watch?v=8DYQQf1KjYo

AI (Haygen) English translation with synchronized lip movement, his voice and appropriate accent: https://twitter.com/aphysicist/status/1 ... 6948907325

candide
Posts: 415
Joined: Fri Apr 08, 2022 9:25 pm
Location: red state America
Contact:

Re: Future of Artificial Intelligence

Post by candide »

Yesterday's Ran Prieur post made a lot of sense to me on this topic.

viewtopic.php?p=284290#p284290

Push-back or other comments probably belong more in this thread than the other.

daylen
Posts: 2466
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Future of Artificial Intelligence

Post by daylen »

If any new technology leads to human transcendence, it will be one that enhances our perception of the living non-human world, and thereby turns our attention outward in a way that was not available to our ancestors.
Seems to have always been the case with tech. Allowing us to see farther or smaller or more comprehensively. AI or IA (intelligence amplification) is a gen 100 candle lighting up the dark corners of our understanding. Helping us to discover more of the universe whether in the space of proteins, planets, batteries, content, concepts, actions, etc. it's all just tokens of prediction and hallucination.

We as humans are highly impressionable stochastic parots that pretend we are the masters of our utterances. Perhaps we are just as lost in the ruliad as the next bot.

candide
Posts: 415
Joined: Fri Apr 08, 2022 9:25 pm
Location: red state America
Contact:

Re: Future of Artificial Intelligence

Post by candide »

daylen wrote:
Fri Jan 19, 2024 12:22 pm
We as humans are highly impressionable stochastic parots that pretend we are the masters of our utterances. Perhaps we are just as lost in the ruliad as the next bot.
Hear hear to that.

But I think Ran is still correct to draw a distinction between Linguistic-utterers who are pointing toward external standards and those that are allowed to go further and further up their own asses. It just seems like LLMs to serve the content esptime is bad for humans and probably bad for the LLMs.

A lot is lost -- and probably key things to non-catastrophic value alignment -- when everything is artificially forced into being seen as fungible, just so it can be commodified.

But, Daylen, you're probably right that Ran isn't seeing how the same tool could just be used in a different way to match the quote you are pulling.

jacob
Site Admin
Posts: 15767
Joined: Fri Jun 28, 2013 8:38 pm
Location: USA, Zone 5b, Koppen Dfa, Elev. 620ft, Walkscore 77
Contact:

Re: Future of Artificial Intelligence

Post by jacob »


daylen
Posts: 2466
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Future of Artificial Intelligence

Post by daylen »

Text to video from OpenAI: https://openai.com/sora

Starting to get quite disruptive to Hollywood (and mimics in other countries). Perhaps Hollywoods' are like quasi-populism centers that will lend way to a more distributed populism that spans socioecononic classes and countries more easily? Combine with YouTube and the barrier to entry for making a film decreases substantially. With a flood of low quality but also some gems here and there that far surpass most blockbusters. Auto translation between any language with synced facial expressions coming along.

bostonimproper
Posts: 574
Joined: Sun Jul 01, 2018 11:45 am

Re: Future of Artificial Intelligence

Post by bostonimproper »

Jesus Christ, it must be $$$$$ to train and run Sora. For reference, estimates for cost to run GPT-4 were on the order of $1MM/day. And OpenAI only made like $200MM in revenue last year. And none of that factors in any of their other expenses like staff, data cleaning, development infrastructure, etc. They must be burrrrning cash.

I imagine this is order(s) of magnitude more expensive to run. Very interested in what the economics of this model today and what their plan is to get this to be performant/cheap enough to be a viable alternative to more manual and rote CGI and filming approaches.

Surely they need to raise more money in the short term, probably another infusion from Microsoft. My guess is that long term GenAI is going to be a loss leader (like rotisserie chicken at the grocery store), getting people in the doors and locked into their partner cloud provider ecosystem. So if you were wary of the power of big tech already, get ready for that to quadruple down.

daylen
Posts: 2466
Joined: Wed Dec 16, 2015 4:17 am
Location: Lawrence, KS

Re: Future of Artificial Intelligence

Post by daylen »

Yeah, it is hard to say where the cards will land. Google released a preview of Gemini 1.5 just hours before sora basically taunting Sam Altman. The corporate games being played are quite interesting to say the least. Gemini is probably going to be used more in educational settings due to its extensive guard railing. OpenAI will probably continue to try appealing to the largest possible audience. Before too long nearly every individual, commune, organization, and government will want to take charge of their own intelligence generation and automated agents and they will not need to start from scratch.

Post Reply