AI is just a tool, and sometimes it can be a tool
https://img.inleo.io/DQmXZV1Uquge5XeqdkzgpcxWAsY2jcug2THYe1jJEewjzPT/ai-generated-8705809_1280.webp
I've have been fascinated with artificial intelligence and machine learning for some time now, long before ChatGPT and the likes got popular. The last few years though have been amazing. Each day something new comes out that just pushes us further into the future.
Artificial intelligence is not as smart as you think
While ChatGPT is amazing, it's really just a glorified auto correct engine. Most of what is being released today is a transformer, aka auto correct.
How ChatGPT works at it's core is has been trained on an incredible amount of text and is then asked to predict the next token. A token is how large language work with text, not the typical words or characters we work off. Think of a token as a group of characters that act as a "meaningful unit" of data.
A lot of times this is individual words, but some parts of words are important as well, for example the word unforgivable, "un" has a significant meaning in the word, and contains a lot of data regarding the context of the input. Another example is the opposite, instead of breaking a word into multiple tokens, a token can be multiple words, such as "New York". The meaning of New and York are considerably different than "New York" together.
Even the same words can be different tokens, for example "The" and "the" have very different meanings as one is the beginning of a sentence or in rare cases part of a name, and the other is likely an article.
ANI vs AGI and ASI
You may not have heard of ANI, but you have most likely heard of AGI or it's more common names "General AI" or "Artificial General Intelligence". I always describe the difference like this. If I taught a computer to play chess, and all it can do well is play chess, this is ANI or Narrow Artificial Intelligence. If I can teach a computer to play Chess and it is able to adapt and play Monopoly, this is a general intelligence. Now if I taught it Chess and it is then able to cure cancer, this is ASI or Artificial Super Intelligence.
A lot of people are saying we are close to or have developed Artificial General Intelligence. I see this talked about a lot and I can see why people think this. I don't believe anything I have seen so far is anywhere near close to AGI. To us, sure it looks that way, but what is actually being done behind the scenes is nowhere near AGI.
Large language models (LLM, ChatGPT, etc) as I said are just auto complete engines trained on so much data it almost can feel like it knows everything or at least can glean it from the training data.
Every day I see something amazing with new models, I see something equally stupid. For example, I was playing with Facebook's new Llama 3 model, and asked it a simple question.
While this in itself isn't a big deal to me, as it is likely not much effort is put into training the model with meta data about itself. Further tests are much more interesting.
When I ask the same question three times, I get three completely different answers with very different (and inaccurate) logic. The question I asked is:
If Sally's brother's each have three sisters, and Sally has four brothers, how many sisters does Sally have? Explain step by step.
This is a very popular test for new models, and in fact it is so popular many models are being trained on it specifically and will sometimes regurgitate an answer from the training data rather than "solve it".
The last part of the prompt "Explain step by step." is very important. This serves two purposes, first it allows us to see how it arrived at its answer. It also will frequently help the model provide a more accurate answer. As you can see, it only did the first.
I can provide examples all day where AI has blown me away with it's answer, and a the same time made me laugh my ass off at how stupid it is. Training, fine tuning, prompting, alignment (aka censorship), and system message all play a big part in how well a model works. Each model can respond very differently to the same prompt, the popular model Claude is well known to require a different approach to prompting for good answers.
AI is just a tool, and sometimes it can be a tool. How you wield and use that tool can dramatically determine your results.
tech new frontier
Post voted 100% for the hiro.guita project. Keep up the good work.
New manual curation account for BEE and Cent
Do you really need to comment you voted for $0.00 on a post?
Sorry if I bothered you. This is a CENT and BEE healing account. Those are the rewards you will receive.
There is no need to tell people you voted them.
Ok
I agree that we are very far from AGI.
These chatbots are incredibly stupid, as well demonstrated by your examples and completely unable to reason.
They are completely unreliable for finding facts because they simply make up crap.
The fact that they lie so much makes them untrustworthy.
The only thing they are really useful for is churning out a different version of text put into them. This can be useful for coding or summarising a longer document.
A lot of components that, potentially, are going to have to enter for AGI to be achieved.
Glorified search is how I look at these chatbots. They are a major improvement over search engines but that is all. Does that mean we will not get the next generation making even greater strides? I wouldn't bet against it.
However, to me, there needs to be a couple more things for AGI. One is spatial since it is not likely to be able to form any general sensibilities without a clear understanding of space. Second, I think some form of symbiotic learning is required. It is one thing to guess the token probabilities. However, unless it can comprehend the similarities (and differences) across many different disciplines, we are just mashing together more data and faster compute.
This could take us far especially as compute increases but will not, in my view, come close to AGI.
the biggest problem that i see as a developer is that chatGPT can never has a context or see what you want, it is like aks to a blind person with super inteligence, he can awser but not ecxlaty what you need :D.
it is a tool and for designs i love it :D
I've played a little with these tools and they can be fun, but they are dumb really. Bear in mind that people were impressed by Eliza 60 years ago!
I'm sure people are working to build some knowledge of how the world works into these models so that their answers make more sense. They are likely to get a lot better. There are probably models for specific purposes that are far better than the free online versions.
I see people slag off the technology, but it's still early days and it won't go away if there is money to be made.
Predictive Text AI is impressive if you wish to write a few pages no one is ever going to read. We have found it useful when travelling to get a quick guide that doesn't necessarily lead you to places that pay to be in a blog.
I find many people who don't speak or write fluent English are using it to pretend they do and to get jobs that are above their pay grade. Employers have to be careful. We also are being inundated with fake AI photos even here on Hive which is impacting the stock photo business. The advances in video and photo AI along with generic writing are going to replace a lot of people in the entertainment and 'art' industry.
Very interesting. I haven't been super impressed with AI yet. I think it feels a bit more like a gimmick at this point, but I think it will get there eventually. These questions you asked are actually pretty interesting. I've never seen something like that before.
How about the other AI codes, such as what is probably being tested in private? Could any of those be getting close?
Maybe. Doubtful. Just larger and more clever training techniques.
How long before we see one that's sentient, do you think?
!PIZZA !ALIVE !LOL
@themarkymark! You Are Alive so I just staked 0.1 $ALIVE to your account on behalf of @ wrestlingdesires. (7/10)
The tip has been paid for by the We Are Alive Tribe through the earnings on @alive.chat, feel free to swing by our daily chat any time you want, plus you can win Hive Power (2x 50 HP) and Alive Power (2x 500 AP) delegations (4 weeks), and Ecency Points (4x 50 EP), in our chat every day.
lolztoken.com
The A-gull and B-gull weren't quite right.
Credit: reddit
@themarkymark, I sent you an $LOLZ on behalf of wrestlingdesires
(6/8)
Delegate Hive Tokens to Farm $LOLZ and earn 110% Rewards. Learn more.
depends on your definition of sentient. I don't think we are anywhere near it. We certainly can simulate it.
Do you think we will get there eventually? I'm referring to the fully self aware type with actual emotions.
!PIZZA !ALIVE !LOL
@themarkymark! You Are Alive so I just staked 0.1 $ALIVE to your account on behalf of @ wrestlingdesires. (10/10)
The tip has been paid for by the We Are Alive Tribe through the earnings on @alive.chat, feel free to swing by our daily chat any time you want, plus you can win Hive Power (2x 50 HP) and Alive Power (2x 500 AP) delegations (4 weeks), and Ecency Points (4x 50 EP), in our chat every day.
lolztoken.com
You are out of jokes for the day!
@wrestlingdesires you can call @lolzbot a maximum of 8 times per day.
Level up by buying more $LOLZ so you can share more jokes per day!
Eventually? Probably.
Why do you ask? Is that something you find preferable?
Why would you wish it to have emotions, for example? How I see it, that is the first reason to not wanting to rely on AI.
What are your thoughts on apps like perplexity, it has internet access and all ways provide references to how it gives the answers.
Perplexity is great, it's the direction search is going moving forward. I don't usually use it, but I have a couple of times.
If you look at Google search you can see they have been moving in that direction as well.
I find myself using it more and more these days, for example if you read some reviews on Amazon and then ask perplexity to filter the comments and find an alternative product. It will recommend a better seller or better version of what you’re looking for that you would have never found yourself.
Just my personal experience with AI in a practical way.
I probably should use it more, but I just feel better being more direct. I don't want to wait around.
$PIZZA slices delivered:
@wrestlingdesires(10/15) tipped @themarkymark (x2)
I haven't done so much with the language ones, but we use MidJourney quite a bit to help with our publicity images for our ensemble concerts. It can be a little bit frustrating at times, but it is sort of interesting to let it run its course and to see what comes out. A lot of people have commented on how interesting our concert images have been.
Recently my wife has taken over that side of things, and after the initial frustration, she has started to get some really nice ones! But there is a lot of trial and error, and we have sort settled on "no humans" as they tend to look weird. Plus, we are usually after more abstract images instead of photorealistic. She also seems to have landed on the idea of just being a gentle guide for the image AI instead of trying to recreate exactly what she has in her mind.
With midjourney and any other image models prompt creation is critical and the difference between a good prompt and a lazy one is massive.
I use copilot on VSS and it is unreal. It's good for the boring health & safety jargon at work as well.
LoL - I did the same and tested an AI-chatbot. Same results you published here.
From the perspective of the human questioner, if he expects to be impressed, he does not bother to ask the same question in repetition. He might tend to take the answer given as true whenever he is not able to find out the answer himself. I agree, you have to ask for the specifics and the path, the chatbot took to come to a result.
To provide my own answer to the Sally question:
Brother 1: 3 sisters
Brother 2: 3 sisters
Brother 3: 3 sisters
Brother 4: 3 sisters
Answer: Sally has 2 sisters.