AI News and Discussion Thread

[UltraMega]

Quote

The rise of AI has elicited fear that the technology will eliminate millions of jobs around the world. The International Monetary Fund this week reported that about 40% of jobs around the world could be affected by the rise of AI.
 

Bill Gates doesn’t necessarily disagree with that notion, but he believes history shows with every new technology comes fear and then new opportunity.
 

“As we had [with] agricultural productivity in 1900, people were like ‘Hey, what are people going to do?’ In fact, a lot of new things, a lot of new job categories were created and we’re way better off than when everybody was doing farm work,” Gates said. “This will be like that.”

This article seems like a good way to kick off this thread. Bill Gates explains how AI will change our lives in 5 years (msn.com)

 

At the title says, this will be a thread to post news about AI and discuss it. Feel free to post any AI related stuff here, or in the news section. Either is fine. I see a lot of interesting news about AI, so I think it will be good to keep a lot of it in one place. 

 

Personally, I think we are somewhere between 6 month to 5 years away from some kind if AI driven robots to enter the market in an impactful way. 

 

 

 

Some interesting talks from tech leaders and insiders about AI:

Bill Gates and Sam Altman (CEO of OpenAI) discussing the future of AI

 

Interview with the CEO of Scale, a company focused on making AI data processing more efficient

 

Jensen Huang of Nvidia on the Future of AI

 

[UltraMega]

Recent news about AI:

 

Google lays off “hundreds” more as ad division switches to AI-powered sales | Ars Technica

Quote

The Information detailed that layoffs would come to Google's Ad division this month. That report said that many of those are being laid off or reassigned because AI is replacing them. Google has been packing Google Ads—its most important product—with tons of generative AI features lately. One is a natural-language chatbot that helps people navigate the large selection of ad products; another is a system that can just make ad assets like images and text on its own based on a budget and goals given by the ad purchaser. Google's generative AI ad system is part of a product called "Performance Max" which works by autonomously remixing and tweaking your ads using the click-through rate as an instant feedback system. Google used to have humans do sales guidance for its products, create art assets, and decide on text and layouts, but now AI can do it a thousand times a second.

 

Artificial intelligence helped scientists create a new type of battery  (sciencenews.org)

Quote

But now a new battery material has been discovered by combining two computing superpowers: artificial intelligence and supercomputing. It’s a discovery that highlights the potential for using computers to help scientists discover materials suited to specific needs, from batteries to carbon capture technologies to catalysts. 
 

Calculations winnowed down more than 32 million candidate materials to just 23 promising options, researchers from Microsoft and Pacific Northwest National Laboratory, or PNNL, report in a paper submitted January 8 to arXiv.org. The team then synthesized and tested one of those materials and created a working battery prototype.

 

How Machine Learning And AI Are Shaping Material Science (forbes.com)

Quote

The advent of computational tools and sophisticated modeling techniques has revolutionized this process. Now, for example, researchers can simulate the behavior of materials under different conditions using advanced software, significantly speeding up the discovery process.
 

This computational approach allows for the exploration of a vast landscape of materials, including those that do not exist naturally, leading to innovations like superconductors, biomaterials and nanomaterials.
 

Moreover, the rise of automation and robotics in manufacturing, underpinned by material science advancements, has enabled the production of materials with precise specifications and in high volumes. This shift boosts efficiency and scalability and opens up new possibilities in product design and functionality.
 

Transitioning to more advanced, technology-driven approaches like integrating machine learning (ML), machine learning operations (MLOps) and large language models (LLMs) is essential to overcome the challenges mentioned above, promising enhanced efficiency, innovation and alignment with sustainability and environmental responsibility. This integration is poised to transform material science, offering new prospects for various industries.

 

NASA accelerates science with gen AI-powered search | CIO

 

Why 2024 Will Be The Year Of Specialized AI (forbes.com)

 

Discoveries in weeks, not years: How AI and high-performance computing are speeding up scientific discovery - Source (microsoft.com)

 

AI-driven drug discovery is poised to boom in 2024 | The AI Beat | VentureBeat

 

 

 

[UltraMega]

Some interesting videos:

 

 

 

 

 

 

 

 

[Slaughtahouse]

Thanks for sharing

 

I came across a video last year which I thought was pretty interesting:

 

 

The presenter is very good at communicating and explaining the whole process. The AI can’t complete Pokemon but they demonstrate how the AI can be trained to learn in a way that is digestible.

 

I have not opened most of those articles but I read about Google layoffs for tomorrow reported by Ars. I’m not entirely convinced that article is accurate. They’re correlating PMAX (AI solution) to LCS (Large Customer Sales) and personally, I think the answer is more simple. Larger customers are spending less (larger trend) and they [Google] need to reorganize their groups and teams internally (trim the fat) to either manage more clients with less or reprioritize smaller clients. I could be entirely wrong but it seems incredibly risky for large clients to switch to AI so quickly at such scale that it’s already resulting in layoffs at Google. Seems sus.

[UltraMega]

37 minutes ago, Slaughtahouse said:

I have not opened most of those articles but I read about Google layoffs for tomorrow reported by Ars. I’m not entirely convinced that article is accurate. They’re correlating PMAX (AI solution) to LCS (Large Customer Sales) and personally, I think the answer is more simple. Larger customers are spending less (larger trend) and they [Google] need to reorganize their groups and teams internally (trim the fat) to either manage more clients with less or reprioritize smaller clients. I could be entirely wrong but it seems incredibly risky for large clients to switch to AI so quickly at such scale that it’s already resulting in layoffs at Google. Seems sus.

Could be a mix of both. Less sales, but also Google wanting to replace humans with AI where possible. I don't think Google is trying to eliminate wholes sales teams overnight, but have an AI that do some of the simpler tasks and thus lessen the need for humans to some degree. 

[Kaz]

This will be a fun topic to follow.

 

Material science has some real potential.   It is a game changer.  My friend who works in AI development (they are working with bees and plant germination), says the problem with AI is no one understands how it generated the answers.  It's a black box that people don't understand.  For stuff where the answer can be checked it's amazing.

 

Chat GPT can generate an answer, but it's also known to lie and make up sources.  It's real goal is to understand languages, but can a statistician ever know if it actually understands?  It can tell you the probability or likelyhood that a certain word will be next.  It doesn't actually know what it's saying.  But hey, if I listen to enough smart people talk I can always just repeat the answer....  The problem is, unless they release their dataset (which they won't because they web crawled it without permission), we don't know how smart those people are when it regurtitates, or just plain makes up answers.

 

https://nytco-assets.nytimes.com/2023/12/NYT_Complaint_Dec2023.pdf

I was trying to copy article 99 to add an excerpt but it doesn't like copying.  It's a blatant example of openAI ripping off an NYtimes article.  It's bad, like I wouldn't have even tried to pull that off in middle school because any teacher would have called it plagurism.

 

Statisticians are scary.  In business stastics we learned how data can be manipulated to show just about anything.  Most of AI development is just statistics.  Where those statistics comes from matters.  What people are trying to show matters.  (I wouldn't trust AMD, Intel, or Nvidia marketing statistics to tell me how good a product is. There's too much variamce between that and independent reviews.

 

OpenAI reached a little hard by not getting authorization to use company datasets  They did it the good old fashioned piracy way, just take it.

[UltraMega]

farm-ng raises Series A funding to scale Amiga robot and AI for small to midsize farms - The Robot Report

WWW.THEROBOTREPORT.COM

---

farm-ng has raised funding to accelerate production of its modular, electric Amiga agricultural robot, which includes...

 

 

 

BMW testing Figure 01 humanoid at Spartanburg automotive plant

WWW.THEROBOTREPORT.COM

---

Figure's humanoid will be tested at BMW Manufacturing's plant in South Carolina. This facility is the largest...

 

 

 

Edited January 20 by UltraMega

[UltraMega]

https://www.msn.com/en-us/money/other/the-us-just-took-a-crucial-step-toward-democratizing-ai-access/ar-BB1hjZFV

 

 

[Slaughtahouse]

AI can now master your music—and it does shockingly well

ARSTECHNICA.COM

---

Suddenly, everyone can master their own music.

 

[Slaughtahouse]

Nvidia’s Chat with RTX is a promising AI chatbot that runs locally on your PC - The Verge

WWW.THEVERGE.COM

---

Chat with RTX is a very early demo app.

 

[UltraMega]

4 hours ago, Slaughtahouse said:

Nvidia’s Chat with RTX is a promising AI chatbot that runs locally on your PC - The Verge

WWW.THEVERGE.COM

---

Chat with RTX is a very early demo app.

 

Hmm so like a local version of copilot. Interesting. Wonder if they will release more local AI stuff like an image generator down the line. 

[UltraMega]

 

 

OpenAI reveals Sora, a very high quality AI video generator. 

 

[GanjaSMK]

I can't wait to see a video of Usher made like a bobble head in his superbowl debut this year. What a half time show it would have been it he'd been a bobble head.... it wasn't a bad show but... with a bobble head it'd have been legendary. 

[UltraMega]

Some thoughts on some of this recent video generation stuff after learning a bit more about it:

 

AI being able to generate these kinds of videos has some implications that go beyond just impressive generation capabilities. It means the AI has a pretty good understanding of the real world and 3D space. The same understanding of the real world that allows AI to make pretty realistic looking videos also means it understands things like physics. 

The ability for the AI to understand physics in this case is an emergent property, meaning it's something the developers didn't set out to do, but happened as a result of training the data to understand the world. 

 

Having an AI that understands the world well enough to generate these videos has a lot of implications. There are more and more rumors and conspiracy theories about OpenAI either already having AGI, or something very close. When I think about what kind of other things an AI with such a deep understanding of the world could potentially be capable of, it starts to seem like AGI rumors could have some merit. 

 

When I try to take things at face value and assume OpenAI doesn't have any big secrets about AGI, the first thing that comes to mind is the 7 trillion figure. Maybe OpenAI isn't realistically anywhere near AGI right now, but they could easily have a very clear path to get there. Perhaps 7 trillion is what Sam Altman thinks it will take to get to AGI. Or perhaps 7 trillion is what he thinks it will take to make AGI accessible in mass and AGI is actually much closer on a small scale. 

 

 

There is so much information to take in with this stuff. I wonder how a person living through the industrial revolution would have felt if instead of happening over roughly a century, it happened much faster. For example, the steam engine was invented in 1712. It took another 50 years for the design to be improved upon to a point where it could be used for a wide range of applications. Steam engine powered trains were not invented until 1804, so it took almost a century to go from steam engines to the first steam train. I'd imagine if you were living through that ear, the progress would have been almost imperceptible in people's general lives. ChatGPT-1 was created near the end of 2018, and I don't think anyone would argue that the advancements from conception to now have been anything short of 'very impressive' at the very least. We are somewhere between the point in which steam engines were invented and mostly used just to pump water out of mines and the point in which it becomes widespread and thus kicks off an industrial revolution, but it's not going to take centuries, or even close. 

Edited February 19 by UltraMega

[UltraMega]

AI solves nuclear fusion puzzle for near-limitless clean energy (msn.com)

 

Quote

The latest research was published in the scientific journal Nature on Wednesday in a paper titled ‘Avoiding fusion plasma tearing instability with deep reinforcement learning’.

“Being able to predict instabilities ahead of time can make it easier to run these reactions than current approaches, which are more passive,” said SanKyeun Kim, who co-authored the study.

“We no longer have to wait for the instabilities to occur and then take quick corrective action before the plasma becomes disrupted.”

 

[UltraMega]

Nvidia CEO predicts the death of coding — Jensen Huang says AI will do the work, so kids don't need to learn (msn.com)