Artificial General Intelligence (AGI) News and Discussions

[weatheriscool]

Intel Says It Will Deliver 5x Increase in AI Performance by 2025
The statement was made during the company's most recent earnings call, indicating it is quintupling down on AI for the future.
By Josh Norem January 30, 2024
https://www.extremetech.com/computing/i ... ce-by-2025

When Intel launched its all-new Meteor Lake CPUs in late 2023, it announced the era of the AI PC had arrived. That's because it has dedicated AI hardware baked into it, and AMD is doing the same with its newest CPUs. Therefore, the race for AI dominance in the CPU world has officially begun, and now Intel is claiming it will achieve a 5X increase in AI performance in the coming years. This signifies that AI performance may become critical when considering a CPU's overall performance.

The announcement about Intel's AI ambitions came in the company's most recent earnings call and was delivered by CEO Pat Gelsinger. On the call, via Tom's Hardware, he stated that this year's Arrow Lake desktop and Lunar Lake mobile platforms will offer triple the AI performance of Meteor Lake. Those platforms will then be followed up in 2025 by Panther Lake on desktop, which will add another 2X uplift for AI. This is the first time we can recall Gelsinger mentioning the AI performance of upcoming platforms on an earnings call, highlighting how important this metric has become for the company's future.

[Tadasuke]

Intel Says It Will Deliver 5x Increase in AI Performance by 2025
The statement was made during the company's most recent earnings call, indicating it is quintupling down on AI for the future.

Intel stated in 2020 that they would deliver 1000x greater general performance by 2025. They also said it was necessary for both VR and AI.

My personal experience is that since 2012 there hasn't been any significant perceptible improvement with using computers.

In 2012, there were already tablets, touchscreen laptops, 240 or 256 GB SSDs, CPUs overclockable to 5 GHz and there was already Samsung Note II with a stylus. Laptops had Ivy Bridge CPUs, Nvidia 600 or AMD 7000 GPUs and 16 GB of RAM. There definitely was less of everything, but software worked similarly to today, WWW worked similarly to today, video games worked similarly to today. No annoying windows asking for permissions or privacy choices, far less ads on YouTube and far less microtransactions.

Artificial Intelligence is just news, discussions, news, discussions, news, discussions, news, discussions and repeat . For the past 25 years or longer. It has become extremely tiresome riding that hype train for the past 25 years. So much time, so little actual improvement. Mostly more spam, more hype, more fake news and more writing or talking about AI-related threats or hopes which don't come true. Why do I even get that AI newsletter on my mailbox?

[Vakanai]

Why do I even get that AI newsletter on my mailbox?

I assume because you haven't opted out yet? There should be a means to be taken off their list so you won't receive it anymore.

[Tadasuke]

I haven't opted out, because I fear potentially missing out on something very important regarding AI. However, with each passing year, my excitement and hopes over AI gradually dwindle.

[wjfox]

[wjfox]

with each passing year, my excitement and hopes over AI gradually dwindle.

I don't know how you could possibly think that, after the past few years of objectively massive AI improvements. It's frankly a bizarre statement.

[firestar464]

also

[Tadasuke]

with each passing year, my excitement and hopes over AI gradually dwindle.

I don't know how you could possibly think that, after the past few years of objectively massive AI improvements. It's frankly a bizarre statement.

I've unfortunately found so far, that AI happened to usually make things more disordered, messy, complicated, convoluted, enigmatic, incomprehensible, perplexing, disconcerting, disorienting, unsettling, distasteful and disconnected than before (so let's say in the year 2015 or in the year 2008).

I myself used to be a teenager with very good grades, relatively high self-confidence and relatively low anxiety, despite not studying that much. I could engage in my hobbies, which gave me reasons to live. Later came anxiety and depression, which lowered my IQ by 20 points, terribly worsened my memory and focus, as well as making my life an unfortunate struggle, in which I don't find enjoyment or peace of mind.

I can totally relate to all the billions of people, who cannot comprehend and get the hang of all of this (exponentially?) growing complexity. I hear lots of people not making sense of all of what is happening. Various addictions, social media, short-video content, attention-economy, news and fake news are causing havoc in people's brains, when many retired people don't even know computer basics. Weird times.

I see only two good solutions:
→ augment humans by some transhumanistic genetic-bio-technology or whatever else works best
→ make things easier, simpler, more comprehensible, more manageable, more intelligible and unambiguous for unaugmented humans
Otherwise only a small minority will be happier and more satisfied.

Last year, one attorney got huge problems by using ChatGPT which gave him completely falsified data, which he presented in the court as true - just an example. Every day I try asking GPT 3.5 and GPT 4 something and it gets things wrong 80% of the time. Voice recognition (mostly from Google) works well only 30-40% of the time. AI is basically a silicon-valley techbros lie to make us even more anxious. :-/

[funkervogt]

[spryfusion]

[firestar464]

I say by the beginning of next year

[spryfusion]

[weatheriscool]

AI outperforms humans in standardized tests of creative potential
https://techxplore.com/news/2024-03-ai- ... ative.html
by University of Arkansas

In a recent study, 151 human participants were pitted against ChatGPT-4 in three tests designed to measure divergent thinking, which is considered to be an indicator of creative thought.

Divergent thinking is characterized by the ability to generate a unique solution to a question that does not have one expected solution, such as "What is the best way to avoid talking about politics with my parents?" In the study, GPT-4 provided more original and elaborate answers than the human participants.

The study, "The current state of artificial intelligence generative language models is more creative than humans on divergent thinking tasks," was published in Scientific Reports and authored by U of A Ph.D. students in psychological science Kent F. Hubert and Kim N. Awa, as well as Darya L. Zabelina, an assistant professor of psychological science at the U of A and director of the Mechanisms of Creative Cognition and Attention Lab.

[Yuli Ban]

Excuse me, it did what?

[firestar464]

Has it merely learned from big brain people online? Or is this something else?

[Yuli Ban]

[Nero]

Absolutely incredible stuff.

[weatheriscool]

Microsoft's small language model outperforms larger models on standardized math tests
https://techxplore.com/news/2024-03-mic ... arger.html
by Bob Yirka , Tech Xplore

A small team of AI researchers at Microsoft reports that the company's Orca-Math small language model outperforms other, larger models on standardized math tests. The group has published a paper on the arXiv preprint server describing their testing of Orca-Math on the Grade School Math 8K (GSM8K) benchmark and how it fared compared to well-known LLMs.

Many popular LLMs such as ChatGPT are known for their impressive conversational skills—less well known is that most of them can also solve math word problems. AI researchers have tested their abilities at such tasks by pitting them against the GSM8K, a dataset of 8,500 grade-school math word problems that require multistep reasoning to solve, along with their correct answers.

In this new study, the research team at Microsoft tested Orca-Math, an AI application developed by another team at Microsoft specifically designed to tackle math word problems, and compared the results with larger AI models.

[weatheriscool]

Balancing training data and human knowledge to make AI act more like a scientist
https://techxplore.com/news/2024-03-hum ... ntist.html
by Cell Press

When you teach a child how to solve puzzles, you can either let them figure it out through trial and error, or you can guide them with some basic rules and tips. Similarly, incorporating rules and tips into AI training—such as the laws of physics—could make them more efficient and more reflective of the real world. However, helping the AI assess the value of different rules can be a tricky task.

Researchers report March 8 in the journal Nexus that they have developed a framework for assessing the relative value of rules and data in "informed machine learning models" that incorporate both. They showed that by doing so, they could help the AI incorporate basic laws of the real world and better navigate scientific problems like solving complex mathematical problems and optimizing experimental conditions in chemistry experiments.

"Embedding human knowledge into AI models has the potential to improve their efficiency and ability to make inferences, but the question is how to balance the influence of data and knowledge," says first author Hao Xu of Peking University. "Our framework can be employed to evaluate different knowledge and rules to enhance the predictive capability of deep learning models."

Generative AI models like ChatGPT and Sora are purely data-driven—the models are given training data, and they teach themselves via trial and error. However, with only data to work from, these systems have no way to learn physical laws, such as gravity or fluid dynamics, and they also struggle to perform in situations that differ from their training data.

[spryfusion]