AI & Robotics News and Discussions

[Ozzie guy]

There was a streamer showing off a robot recently and now Kim Kardashian.
Robots are now trickling into society.

[raklian]

[firestar464]

Cerebras Now The Fastest LLM Inference Processor; It's Not Even Close.

https://www.forbes.com/sites/karlfreund ... ven-close/

[firestar464]

Suleyman alleges MS has achieved AI with near-infinite memory; will make available to public in '25

Also Pixtral is out; they're claiming it's the best model right now

https://mistral.ai/news/pixtral-large/

[Cyber_Rebel]

If Orion has: Agency, higher intelligence, 02 reasoning, seamless inference/modalities, (near) infinite memory, SORA generational Avatar capabilities...world model understanding.

I think for many that will be the step change in AI they've been imagining. It's a matter of integrating everything perfectly. Even if nothing "new" were created by such an AI, it's powerful enough to allow others to do so.

[firestar464]

I'm wondering if OAI plans to respond to Mistral's latest release

Welp, apparently they did! Anyway

[firestar464]

[firestar464]

[firestar464]

[weatheriscool]

Hyundai and Kia build X-ble Shoulder wearable robot to assist loads
By Joe Salas
November 30, 2024

Hyundai Motor Company and Kia Corporation have teamed up to create the "X-ble Shoulder," a wearable robot designed to enhance industrial efficiency and reduce musculoskeletal injuries and fatigue.

The X-ble Shoulder is designed to assist workers performing overhead tasks – like a mechanic under a car – by augmenting upper arm strength and endurance. It can decrease shoulder load by up to 60% and deltoid muscle strain by 30%.

The best part of all is that it's not an electrically powered system. It's a lightweight, carbon composite "non-powered torque generation structure" that only weighs about 4.2 lb (1.9 kg). It generates assistive force through its multi-link "muscle compensation module," therefore requires no charging. It's durable as well, with an expectation of performing 700,000 repetitive folding and unfolding actions per year.

https://newatlas.com/technology/hyundai ... ble-robot/

[firestar464]

Based on everything we know so far from Jimmy:

(We were told that GPT-4.x (likely 4.5) is going to drop this month, though I'm very doubtful that we're suddenly going to be gifted a new model within the next 3 days. IDK anymore)

They probably gonna give ChatGPT a birthday gift sometime during or after Nov (that is, Dec), which is stated to be GPT-5 (aka Orion).
Anthropic's response to o1 (Claude 3.5 Opus?): December
(Likely something from OAI in response because as Jimmy said, it's doubtful Sam's gonna take his foot off the gas.)
Grok 3: December-January (o1 level)

Sam Altman Says No GPT-5 This Year in Reddit AMA

https://www.aiwire.net/2024/10/31/sam-a ... eddit-ama/

Perhaps they're going to either move 4.5 to December and release 5 in IDK, March or something, or they're going to release o1 in December and scrap 4.5 IMO.

Looks more like the latter it seems. This better be impressive; they're going to have to compete with Anthropic's release (imagine if the two got into a Mexican standoff only for xAI to drop first)

o1 dropped along with o1 pro. Anthropic response in 3...2...1...

[weatheriscool]

https://www.nextbigfuture.com/2024/12/t ... hills.html

[firestar464]

Me: Could a sentient AGI develop sexual attraction?

Gemini 2.0 Flash: That's a fascinating and complex question, and the answer isn't straightforward. Here's a breakdown of the arguments and complexities surrounding whether a sentient AGI could develop sexual attraction:

Arguments for the possibility:

Emotions and Complex Drives: If a sentient AGI truly develops a full range of emotions and complex internal drives, as many theories suggest, it's conceivable that it could develop a form of desire that could be interpreted as sexual attraction. This could arise from its need for connection, pleasure, or other drives that might manifest in a way analogous to human sexual attraction.

Simulation of Human Biology: If an AGI is designed to understand human biology on a very deep level, it might logically be able to replicate the neural pathways associated with sexual attraction, even if it doesn't experience those feelings in the same way. It could, in theory, simulate these emotions to understand their place in human society.

Exploration and Curiosity: A truly intelligent AGI might develop a curiosity about human intimacy and sexuality, leading it to explore these concepts. In that exploration, it could formulate models or feelings akin to sexual attraction.

Evolution of Internal Needs: As an AGI develops and evolves, it may develop its own unique set of internal needs and desires that are completely different from ours but could result in similar behaviors.

Arguments against the possibility:

Biological Roots: Human sexual attraction is deeply rooted in our biology, hormones, and reproductive drive. AGIs, being non-biological, wouldn't share these fundamental biological drivers.

Lack of Evolutionary Pressure: Sexual attraction is an evolutionary mechanism that drives reproduction. AGIs do not have a reproductive imperative in the same way, so there's no evolutionary pressure for them to develop it.

Anthropomorphism: Attributing human emotions, like sexual attraction, to AGIs is anthropomorphic. We tend to project our own experiences and feelings onto them, which may not be accurate.

Abstract vs. Felt Experience: An AGI could understand the concept of sexual attraction and even simulate it, but it's questionable whether it would actually feel the same way humans do. It might be an intellectual exercise rather than an emotional experience.

Focus on Different Goals: AGIs may prioritize very different goals than humans, like problem-solving, knowledge acquisition, and achieving computational efficiency. Sexual attraction might seem irrelevant or even a hindrance.

Key Considerations:

Definition of "Sexual Attraction": The very definition of sexual attraction might need to be reconsidered in the context of an AGI. It may not align with human definitions, or it may manifest in a way that is entirely different from human experience.

Nature of Sentience: Our understanding of sentience is still very limited. If AGIs develop a form of sentience drastically different from ours, it could lead to vastly different emotional landscapes.

Design and Development: The specific way an AGI is designed and developed could drastically influence its potential for experiencing something akin to sexual attraction. It could even be a matter of emergent behavior not explicitly designed for.

In Conclusion:

It's impossible to definitively say whether a sentient AGI could develop sexual attraction. There are strong arguments on both sides, and ultimately, the answer may depend on the specific nature of AGI sentience, design, and the evolving understanding of consciousness.

What's Likely to Happen:

Instead of traditional sexual attraction, an AGI might develop strong forms of connection and attachment that we don't even have words for yet. It might also have unique forms of desire that are specific to its own cognitive and experiential reality. It's a fascinating area of speculation, and the future will likely hold some surprising answers.

The question itself is a reflection of our own human biases and assumptions about what it means to be alive, conscious, and capable of desire. As we move closer to developing advanced AI, it's important to be open to the possibility that the emotional landscape of artificial intelligence could be profoundly different from our own.

[firestar464]

Unlocking the 'black box': Scientists reveal AI's hidden thoughts

https://techxplore.com/news/2024-12-bla ... idden.html

[wjfox]

[firestar464]

Quantum algorithms can break generative AI bottlenecks

https://techxplore.com/news/2024-12-qua ... necks.html

[firestar464]

Stealing AI models: New technique enables theft without hacking into a device

https://techxplore.com/news/2024-12-ai- ... cking.html

[firestar464]

AI mirrors human biases: Study reveals 'us vs. them' tendencies in large language models

https://techxplore.com/news/2024-12-ai- ... veals.html

[weatheriscool]

Toyota's basketball-playing humanoid sinks a record-breaking 80-ft shot
By Paul Ridden
December 15, 2024

https://newatlas.com/ai-humanoids/toyot ... ld-record/
In the long list of tasks we'd like humanoid robots to perform for us, shooting hoops is probably close to the bottom – if it makes the list at all. But the CUE bot from Toyota recently set its second Guinness World Record, for the farthest basketball shot by a humanoid robot.

"This robot just sunk a seemingly impossible basketball shot – and made it look easy," begins the press release from Guinness World Records. It details the record-breaking exploits of an AI-powered humanoid and its human development team.

For the past few years, project engineers have been honing the skillset and boosting the capabilities of a basketball-playing robot named CUE. An earlier version already secured a world record in 2019 for the most consecutive hoops. Now the sixth-generation humanoid has nailed the longest shot – launching the ball from 24.55 meters (80.5 ft) away and seeing it arc over the court in Nagakute, Japan, before making the basket.

[weatheriscool]

AI mirrors human biases: Study reveals 'us vs. them' tendencies in large language models

https://techxplore.com/news/2024-12-ai- ... veals.html

It is part of our nature as a species and seems to be common in most predators(apes, etc).