4·
11 months agoHe meant Lexus but he ain’t know it.
- 0 Posts
- 58 Comments
Joined 1 year ago
Cake day: July 23rd, 2023
You are not logged in. If you use a Fediverse account that is able to follow users, you can follow this user.
2·11 months agoSince the forces that determine policy are largely tied up with corporate profit, promoting the interests of domestic companies against those of other states, and access to resources and markets, our system will misuse AI technology whenever and wherever those imperatives conflict with the wider social good. As is the case with any technology, really.
Even if “banning” AI were possible as a protectionist measure for those in white-collar and artistic professions, I think it would ultimately be unfavorable with the ruling classes, since it would concede ground to rival geopolitical blocs who are in a kind of arms race to develop the technology. My personal prediction is that people in those industries will just have to roll with the punches and accept AI encroaching into their space. This wouldn’t necessarily be a bad thing, if society made the appropriate accommodations to retrain them and/or otherwise redistribute the dividends of this technological progress. But that’s probably wishful thinking.
To me, one of the most worrying trends, as it’s gained popularity in the public consciousness over the last year or two, has been the tendency to silo technologies within large companies, and build “moats” to protect it. What was once an open and vibrant community, with strong principles of sharing models, data, code, and peer-reviewed papers full of implementation details, is increasingly tending towards closed-source productized software, with the occasional vague “technical report” that reads like an advertising spiel. IMO one of the biggest things we can lobby for is openness and transparency in the field, to guard against the natural monopolies and perverse incentives of hoarding data, technical know-how, and compute power. Not to mention the positive externality spillovers of the open-source scientific community refining and developing new ideas.
It’s similar to how knowledge of the atomic structure gave us both the ability to destroy the world, or fuel it (relatively) cleanly. Knowledge itself is never a bad thing, only what we choose to do with it.
I take your point, but in this specific application (synthetically generated influencer images) it’s largely something that falls out for free from a wider stream of research (namely Denoising Diffusion Probabilistic Models). It’s not like it’s really coming at the expense of something else.
As for what it’s eventually progressing towards - who knows… It has proven to be quite an unpredictable and fruitful field. For example Toyota’s research lab recently created a very inspired method of applying Diffusion models to robotic control which I don’t think many people were expecting.
That said, there are definitely societal problems surrounding AI, its proposed uses, legislation regarding the acquisition of data, etc. Often times markets incentivize its use for trivial, pointless, or even damaging applications. But IMO it’s important to note that it’s the fault of the structure of our political economy, not the technology itself.
The ability to extract knowledge and capabilities from large datasets with neural models is truly one of humanity’s great achievements (along with metallurgy, the printing press, electricity, digital computing, networking communications, etc.), so the cat’s out of the bag. We just have to try and steer it as best we can.
1·11 months agoI have to disagree about that last sentence. Augmenting LLMs to have any remotely person-like attributes is far from trivial.
The current thought in the field about this centers around so-called “Objective Driven AI”:
in which strategies are proposed to decouple the AI’s internal “world model” from its language capabilities, to facilitate hierarchical planning and mitigate hallucination.
The latter half of this talk by Yann LeCun addresses this topic too: https://www.youtube.com/watch?v=pd0JmT6rYcI
It’s very much an emerging and open-ended field with more questions than answers.
In a sense… yes! Although of course it’s thought to be across many modalities and time-scales, and not just text. Also a crucial piece of the picture is the Bayesian aspect - which also involves estimating one’s uncertainty over predictions. Further info: https://en.wikipedia.org/wiki/Predictive_coding
It’s also important to note the recent trends towards so-called “Embodied” and “4E cognition”, which emphasize the importance of being situated in a body, in an environment, with control over actions, as essential to explaining the nature of mental phenomena.
But yeah, it’s very exciting how in recent years we’ve begun to tap into the power of these kinds of self-supervised learning objectives for practical applications like Word2Vec and Large Language/Multimodal Models.
Many modern theories in cognitive science posit that the brain’s objective is to be a kind of “prediction machine” to predict the incoming stream of sensory information from the top down, as well as processing it from the bottom up. This is sometimes referred to through the aphorism “perception is controlled hallucination”.
Get Pudgie Walsh on the horn. He’ll straighten this out.
The Japanese SCMaglev only has the cooling stuff on the train, not along the entire length of the track.
And I think there is a “high-temperature SC Maglev” in development in China too.
Why no superconducting maglev tho?
2·1 year agoThat’s useful to know that it at least mostly works. I should really try it out with my Thrustmaster T300, I could be pleasantly surprised. I use an Oculus Quest 2 headset, which requires Meta’s app to run on Windows, so not sure how that would pan out.
If I could one day be playing BeamNG, with my FFB wheel, in VR, on Linux - I will have truly attained nirvana.
TBF I haven’t actually tried Asetto Corsa with my steering wheel, or XPlane with my VR headset on Linux yet I just assumed it wouldn’t work. As soon as they do, I can’t wait to shitcan Windows forever.
Dig your own grave and save!
7·1 year agoLocalsend - it’s like airdrop but for android, linux, windows, and mac.
1·1 year agoDeontay needs to get his brother
ML has already had a huge impact on the world (for better or worse), to the extent that Yann LeCun proposes that the tech giants would crumble if it disappeared overnight. For several years it’s been the core of speech-to-text, language translation, optical character recognition, web search, content recommendation, social media hate speech detection, to name a few.
Some of the current thought on shortcomings of LLM capabilities actually takes influence from human cognitive science, and what can be learned from those with neurological impairments. It’s thought that human language abilities are strongly dissociated from other reasoning abilities because individuals with aphasia can lack the ability to speak or comprehend language, yet be able to solve mathematical problems, engage in logical reasoning, enjoy music, categorize objects and events, etc.
It’s shown that LLMs develop a crude world model for performing reasoning tasks, yet it’s inextricably tied up with their language functionalities (since they are ONLY language based). The hope for future research is to develop AIs with world models and planning faculties that are decoupled from the language analysis module, which would mitigate hallucination and aid in interpretability.
Free speech online doesn’t even seem to be a particularly well-defined concept. Those who extol it the loudest are often looking to have the millionth “good faith discussion” about The Bell Curve, or use slurs as “just a joke”, or promote a “dating and lifestyle coaching” business to teenage boys. If all they want is carte-blanche to say absolutely anything without being censored, I guess they only need to spin up a web server of their own, or run a lemmy instance. But what they actually want is to bypass the moderation rules on widely-used platforms and shit on the social contract. It’s the same reason they don’t show pornography, snuff footage, or other damaging content on television.
4·1 year agoMaybe he meant that holocaust of 2pac they did at cochella.
But when it comes to battery power tools, you have to pick a brand and stick with it, unless you’re John D Rockefeller with 6 types of charger and a billion battery packs.