But what is innovation? Neither Newton nor Einstein could have 'innovated' without there being an existing meme pool of ideas that was ready to be innovated in. There is no evidence that innovation is anything other than memetic copying with a bit of random mutation. Likewise, all it takes for an LLM to be creative is a little bit of random error from the 'known facts'.
-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Master AI (Artificial Intelligence) Discussion/News Thread
- Thread starter amirm
- Start date
But what is innovation? Neither Newton nor Einstein could have 'innovated' without there being an existing meme pool of ideas that was ready to be innovated in. There is no evidence that innovation is anything other than memetic copying with a bit of random mutation. Likewise, all it takes for an LLM to be creative is a little bit of random error from the 'known facts'.
It's usually that random mutation that is very novel, and is therefor innovation, often not trivial as suggested. Not sure if the current AI is thinking "out of the box" yet and agnostic about when it will, if ever. Agree about Newton and Einstein, they were genus applied mathematicians.
Yes, I've never really thought we have a metaphysical place in the universe. A physical one is plenty for me ...
Agreed. The technology itself is neither good nor bad. But the megalomaniac tech-bro loons that control it coupled with naïve users who give it far more credence than is warranted – that can and will be problematic.
That may be true, but THEY came up with the discontinuity, the breakthrough that changed the status quo. They didn't merely extrapolate from an existing data pool.
"Innovation" very often is just reuse of existing stuff. "Inference" describes that type of innovation.
I didn't mean to suggest any of this is trivial. But mutations are small differences that with time, and many iterations, become big differences. That's how we got from single celled organisms to oak trees, blue whales and humans. And, as noted above, that's why we didn't have smartphones in the paleolithic: no 'prior art' ...
Okay but there are many instances of step functions or mistakes that happen to pay off as innovations with major changes.
Well, that's exactly what I think they did do. To put it another way, both Newton's and Einstein's innovations would have almost certainly appeared within a decade or two had they themselves not published when they did. It was 'in the air' so to speak.
To get back to AI, I don't for a moment think we will have 'AGI' by 2030 like the tech-bros claim. Mainly because there is no such thing as AGI. But, by small increments (innovations) we will, sooner rather than later, achieve something immensely powerful and indistinguishable from human ingenuity and creativity. Probably not in my lifetime. Whether it will be 'truly' creative or 'truly' innovative is neither here nor there, as long as it's useful ...
The process of natural evolution and the process of AI inferencing are extremely different in their very nature.
Natural evolution is chaos theory in effect. That is the utter opposite of what you want with AI, which is supposed to be extremely predictable and deterministic. And AI doesn't evolve outside of its designed purpose. If you train a CNN (convolutional neural network) to recognize hamsters in pictures, it has zero ability to recognize a rabbit unless you train it from scratch.
coonmanx
Addicted to Fun and Learning
Would AI ever tell you that it likes bacon. I don't think so. Therefore AI is not real in any way. Because even if you do not eat bacon, you know that people like it....
LOL.
Watch the Dog Whisperer and learn... Don't just vacuum the internet...
LOL.
Watch the Dog Whisperer and learn... Don't just vacuum the internet...
But the process of AI inferencing getting better over time is a gradual evolutionary process. Re. your claim that AI 'cannot be creative': There was a time when humans weren't creative, or even human for that matter. Then we got a little bit more creative and more human. Then a little bit more and so on. We cannot draw a hard and fast line where 'humanity' began or where 'creativity' began. Likewise for AI 'creativity'. Even if we accept that AI is not creative yet (which I personally would question) there is nothing intrinsically to stop it becoming so with time. The biggest hurdle is probably the energy to power it ...
As Pope Leo wrote in his recent encyclical on AI systems (probably abetted by Anthropic's Christopher Olah): " ... all of us, including those who design them, possess only a limited understanding of their actual functioning. Indeed, current AI systems are more “cultivated” than “built,” for developers do not directly design every detail, but instead create a framework within which the intelligence “grows.” As a result, fundamental scientific aspects — such as the internal representations and computational processes of these systems — remain, at present, unknown."
I don't think it's "supposed" to be anything. Who says is it supposed to be predictable? If I was an AI developer ( to be clear, I'm not, nowhere near it...) surely I would want to explore what it could do without these constraints (if indeed they actually exist anymore)
I reckon it will have this ability. Indeed I reckon it probably already does
When AI Teaches Itself: The Breakthrough of Zero-Data Learning
Self-Improving AI Systems That Need No Human Training
www.theaugmentededucator.com
- Joined
- Jul 21, 2019
- Messages
- 3,543
- Likes
- 5,257
Unnamed large company gives employees unlimited access to Claude, racks up $0.5 billion USD in charges in a single month.
https://newsinamerica.com/en/business/2026/ai-sticker-shock-hits-corporate-america/ (originally appeared on Axios)
https://newsinamerica.com/en/business/2026/ai-sticker-shock-hits-corporate-america/ (originally appeared on Axios)
Hysterical.
If true - (sounds like an anti-AI hit job slop piece) - all this would be illustrating is how poorly that companies decision making, management, training and implementation people are - from top to right before the end user. This would be akin to "company gave employees unlimited access to corporate spending account and their fleet of private jets to go expand client base and were shocked at the expense report".
- Joined
- Mar 1, 2021
- Messages
- 3,146
- Likes
- 6,733
I’d rather a javelin struck the bubble
Everybody is forgetting AltaVista, Lycos, Ask Jeeves, Excite, HotBot WebCrawler Dogpile and dial-up speeds. The data center thing is a way to lock down proprietary data which will come from personal search data that can be resold. IMO successful AI will be in instantaneous data usage from large but much smaller databases in transportation, logistics, automation, robotics, identification of all sorts, etc not just faster more detailed searches, programing and customer service. If the gloomers are right jobs will be lost and a global recession will hit resulting some large cheap sources of electricity as the looser's will need to recoup their investments.
Yes but the actual issue... I fear tremendously what AI may bring about. The issue is the noise. It is the amount of top contributers who are spurting out their feelings as facts or referencing missive as dismissal of the contributers providing real and helpful sharing of successes and how others can benefit with some tips on how to get the most out of AI.. They are not debating they are spitting out noise and snide remarks and giving each other thumbs up. Now what I realize my shock at this is I mistook them as moderators rather than mass posters. My desire is to here all sides and to debate intelligently and respectfully. Straw man arguements both sides to turn away. if your post is I don't care I will never... Or AI is x case closed and it is a experience you express as universal truth you are creating noise and then playing the victim card of being dismissed when you are being dismissive.
I'm not professionally into AI, but see two directions of use:
One is within conversation to anyone on the web,
the other one is to use it internally in industry process optimization.
The first is, as can been seen at schools today, unhelpful,
the second might be helpful.
It all depends ...
One is within conversation to anyone on the web,
the other one is to use it internally in industry process optimization.
The first is, as can been seen at schools today, unhelpful,
the second might be helpful.
It all depends ...
Chat based AI has several modes and this is important to understand.
Standard free mode is optimized to be fast to response. It looks at cached data. It also can optimize to how you query and interact. If you are asking about the nba it responds like a fan. If ask about Taylor Swift it interacts like social media. If you ask a technical question it tries to answer technically. If you don't question it it will respond.more sure of itself. It is not acting as a scientist it is acting as a pleaser or more precisely applying the probable next word to that of who it thinks it interacting with.
The free version also allows you to switch to a more advanced mode but for a daily limited interaction. It is optimized, for more current and potentially more accurate responses.
Paid tiers will provide both modes, as well but you want to set the default to the advanced mode which you now have unlimited cycles. You also have access to more, advanced models potentially. Read the description on what you get. This will provide greater accuracy and greater precision.
But chat AI is not a single function engineered precision machine and understand what it excels at. And it as with all responses you need to determine what level of confidence you need based on how, critical your question subject and consequences is
You can also offer prompts to set guardrails. I put in things like no use of reddit. Go to vendor and expert sites only. It learns what you prefer but you can also say look at, ASR and other science based sites. A great prompt I just thought of is give me the science engineering response.. And the the Audiophile based response. Then give me your AI advice on what aspects to pay attention to or dismiss or at least question.
You can do these things in free tier just it may be capped but most interactions will not hit the cap and if you do you can pickup where you left off 24 hours later.
If you find it still sucks you may indeed found something it sucks at or you may want to rethink reexaminea better way to phrase your query.
Standard free mode is optimized to be fast to response. It looks at cached data. It also can optimize to how you query and interact. If you are asking about the nba it responds like a fan. If ask about Taylor Swift it interacts like social media. If you ask a technical question it tries to answer technically. If you don't question it it will respond.more sure of itself. It is not acting as a scientist it is acting as a pleaser or more precisely applying the probable next word to that of who it thinks it interacting with.
The free version also allows you to switch to a more advanced mode but for a daily limited interaction. It is optimized, for more current and potentially more accurate responses.
Paid tiers will provide both modes, as well but you want to set the default to the advanced mode which you now have unlimited cycles. You also have access to more, advanced models potentially. Read the description on what you get. This will provide greater accuracy and greater precision.
But chat AI is not a single function engineered precision machine and understand what it excels at. And it as with all responses you need to determine what level of confidence you need based on how, critical your question subject and consequences is
You can also offer prompts to set guardrails. I put in things like no use of reddit. Go to vendor and expert sites only. It learns what you prefer but you can also say look at, ASR and other science based sites. A great prompt I just thought of is give me the science engineering response.. And the the Audiophile based response. Then give me your AI advice on what aspects to pay attention to or dismiss or at least question.
You can do these things in free tier just it may be capped but most interactions will not hit the cap and if you do you can pickup where you left off 24 hours later.
If you find it still sucks you may indeed found something it sucks at or you may want to rethink reexaminea better way to phrase your query.
Last edited:
Unwatching this topic. Have fun!
Similar threads
