Tom Scott: I tried using AI. It scared me. [video]
youtube.com
youtube.com
ChatGPT is a proof-of-concept. There are agents far, far more capable than it on their way, and it is impossible to know for sure how they will impact the world at large. We are observing science fiction become science fact and it is really unnerving at times.
Tom is a smart and thoughtful guy, and I think he's right on the money with this video.
I don't find very much substance in this argument, given how impossible it is to consistently define "new" and "idea" in this context. At the end of the day, ChatGPT demonstrates that AI models are making real progress towards convincing mimicry of human behavior. How it does it is of no concern to me, what is concerning is the fact that it does it at all.
Babies learn language by being exposed to many many examples of it, why shouldn't the same work for artificial intelligences? The concept of "new ideas" is also flawed, as nothing is truly new—our actions, feelings, thoughts, desires, and ideas are all recombinations of our experiences. Even the ideas that seem to come out of nowhere must come from somewhere.
Im not saying it's not the future, it is, we just need the hardware, once there's a hardware cheap enough that replicates what an human does with his hands, then the days working at a factory will be over. The current AI's are perfect for copycating behavior.
«just a text generator on steroid » is meaningless unless you define what « on steroid» means. It might as well describe a novelist’s activity or a programmer.
The ability to understand enough of the relationship between words and their associated concepts in chatgpt is good enough to produce meaningful and sometimes correct answers to a huge number of questions. This is far beyond « random », and can easily be described as an « intelligent » capacity.
Chaos theory just happens to be deterministic. And life is most assuredly chaotic. In fact it was equations related to populations of living things that lead to the discovery of chaos theory. (equations wrt the weather also helped) [1]
So I'd make your statement:
Our thoughts, ideas, words, and actions are really really irreducibly new. They are (deterministic-)chaotic transformations based on our genetics and sensory input.
Every web-app or scientific application I've written is merely "rewriting" and stitching together other techniques, libraries, and patterns. I haven't broken any new ground on data encoders/process handling/data analysis, but my work had value.
In this scenario, if ChatGPT provides enough value to reduce my workload and allow me to allocate my effort elsewhere, why is it not a valuable tool, regardless of whether it "creates anything new"?
This same argument can be made against the concept of an IDE. My IDE doesn't create anything new. I used to use Vim with a number of plugins, a terminal multiplexer, and an auto-completion engine.
Why would <insert IDE of choice> improve my workflow? It doesn't add anything new to my existing workflow. But after migrating to my current IDE, the convenience it brought me was significant, even if it didn't add anything new to my setup.
could you make some that might be useful to aliens
Sure! Here are a few SaaS ideas that might be useful to extraterrestrial beings:
A intergalactic communication platform, allowing aliens from different planets to exchange messages and share information.
A virtual reality platform for exploring other planets and galaxies, providing a way for aliens to experience and study new environments from a safe distance.
A space weather forecasting software, providing up-to-date information on solar flares, coronal mass ejections, and other space weather events that could affect intergalactic travel and communication.
An artificial intelligence-powered resource management system, allowing aliens to optimize resource usage and minimize waste on their home planets.
A platform for remote monitoring and control of extraterrestrial mining operations, including real-time data analysis and reporting.
A language translation software that can translate between different extraterrestrial languages and dialects.
A virtual market place for intergalactic trade, facilitating commerce between different alien species.
These are just a few examples of SaaS ideas that could be of use to extraterrestrial beings. The key is to think about the unique needs and challenges that aliens might face in their everyday lives, and offer solutions that address those needs in innovative and effective ways.
https://www.youtube.com/watch?v=419no81su64
I'm pretty into cocktails and I've never heard of the ones it came up with. Admittedly, the world of cocktails is huge and every single cocktail is 'just a riff' on some other one at the end of the day. In short, these drinks are new.
Futhermore, to me it shows 'learned' innovation on the AI's part. Like, the AI got ~95% of the way there, just what a human would do when trying to create and refine drinks.
Does this answer your question?
The danger of ChatGPT is that it requires little effort to generate massive amounts of content, but much more effort to debunk and fact check. It will become increasingly difficult to separate the signal from the noise as the internet and other places are flooded with auto-generated content.
Personally I'm not concerned about ChatGPT becoming some sort of "intelligent overlord". However I am concerned about how other humans will wield this increasingly powerful tool.
I meant that could be used by AI chatbots
Oh, I see! Here are a few SaaS ideas that AI chatbots could use:
A natural language processing (NLP) platform to improve the accuracy and efficiency of chatbot conversations.
A platform for managing and analyzing chatbot data, providing insights into customer interactions and feedback.
A platform for integrating chatbots with multiple messaging channels, including popular social media platforms, messaging apps, and websites.
A platform for building and training custom AI chatbots, allowing businesses to create their own AI-powered customer service and support solutions.
A platform for managing and deploying chatbots at scale, providing tools for monitoring and maintaining multiple chatbots across different channels.
A platform for automating chatbot deployment and updates, simplifying the process of deploying and maintaining chatbots across different channels.
A platform for enhancing the user experience of chatbots, including tools for personalization, gamification, and voice and text-based interactions.
These are just a few examples of the types of SaaS solutions that could be used by AI chatbots to improve their functionality and user experience. By leveraging the latest advances in AI and NLP technology, these solutions could help chatbots better understand and respond to customer needs and preferences.
Sure, here are some unique and unusual personas that could serve as inspiration for building new and innovative apps:
Conspiracy theorists - An app that curates a personalized feed of conspiracy theories, and provides tools for users to explore and validate their claims.
Astrology enthusiasts - An app that combines astrological analysis with personalized daily horoscopes and suggestions for self-improvement.
People who love to nap - An app that helps users track their napping habits, suggests optimal napping times based on their sleep patterns, and provides guided meditations and ambient sounds to help them fall asleep faster.
People who are fascinated by death - An app that allows users to explore different cultural beliefs and rituals around death, and provides tools to plan and organize their own end-of-life wishes.
People who are into sensory deprivation - An app that provides a virtual reality experience for users to explore different states of sensory deprivation, such as floating in a sensory deprivation tank or experiencing total darkness.
People who are into near-death experiences - An app that collects and curates personal accounts of near-death experiences, and provides tools for users to reflect on and analyze their own experiences.
People who are into cryonics - An app that provides information and resources about cryonics, and helps users plan and prepare for their own cryopreservation.
I hope these ideas spark your imagination and inspire you to build something new and unique!
I think we're right before the hockey stick, just like the days of Napster.
To use another analogy - The Industrial Revolution didn't make it possible to make screws, it made it possible to make a box of screws that you could sell for a dollar. Previously it took a skilled craftsman and time. No two were alike, and they tended to be individually stamped with a number so you could unscrew/rescrew things when you took something apart.
Someone in 1810 wouldn't want to buy cheap screws, the way forward would be able to own a lathe, planer, and make your own screws, and other hardware.
Similarly, my strategy right now is to try to find stuff I can use on my own computer, not something that Google can kill on a whim.
I've played with Stable-diffusion-ui, and found that for me, the main use would be to transform a rough (VERY rough) sketch of something into well illustrated picture.
I'll start playing with ChatGPT and the like when I can run them on my own hardware.
It seems closer to this: https://xkcd.com/605/
I'll concede there have been large breakthroughs, but to extrapolate from that there will be even larger breakthroughs (without a really solid account of why that would be) seems very sketchy. I get that the people who are trying to sell the idea to investors are pitching this line, but applying a modicum of critical thinking, that's relatively rarely how things work. Returns tend to diminish with the amount of effort put in, not exponentially increase.
I personally believe in the long-term vision of AI/ML/RL. Progress will continue to be made in the long run. However, we don't know what we don't know. We may hit a wall in a few years that takes decades to overcome. On the other hand, perhaps the current rate of innovation will continue for the forseeable future. Time will tell.
I think we may be on the S part of the curve, but thanks to processing concerns we will be in for veeeeeerrryyyy gentle slope on that S. Capital is getting more expensive, processors getting less fast, and in general tech is burdened with a slowdown.
We are in for a revolution, I believe, but one that will be a bit slower than the internet. More akin to maybe the revolution of chemical engineering, something that lasts a lifetime instead of a decade.
What makes you think that?
Although neural networks are capable of any form of reasoning or understanding and human language is a very strongly diverse field so I do believe LLMs are a strong contender for the production of general AI.
Lots of data.
Easy to train thanks to text being very simple.
Easy to express just about any concept or information.
Requires some amount of domain knowledge to make good prediction of what comes next.
ChatGPT is just a really, really, REALLY good search. Things that used to take multiple google searches now take one query. This will definitely disrupt quite a few things and render some obsolete, but generally the effect will just be making things more efficient.
There are three things IMO that is going to make us start shooting up the final part of S curve.
First is efficient on the fly learning. If you think about how humans learn, we don't retrain ourselves every time we recieve a new piece of information that doesn't go contrary to what we know - a neural network however needs to be iteratively trained on it, with weight changes in most every single one of its parameters. We need models like Chat GPT to be presented with new text, and not only "remember" it, but also contextualize it.
Second is the contextualization mechanic itself. Its well known that generative models are statistical at their core, ChatGPT is wrong in a lot of cases. There needs to be some way to encode truth into the model.
Third is applying neural nets towards generating neural net architectures (or more specifically optimized compute graphs). The tricky part here is to determine what the loss function is (you wouldn't want to generate a net, train said net, evaluate its performance, then feed that back into the original generative net).
Koboldai
I've noticed this too with friends and coworkers. People that I would think would have no business knowing about chatGPT are already using it in their work and their lives.
Almost axiomatically, that means that we have to be in the 'mass adoption' phase to some degree.
But I don't agree. I think we're in the early shallow phase still.
'The Kids' aren't the ones using it. I'm not being left behind by the 'hip new thing' like I should be. To me, that means that the nerds are still gawking about with it, trying on new things. Once 'The Kids' get a hand on it and use it in the innocent and unabashed way that they always do, then I can say we're in the early-mid phase. And by 'The Kids', I mean not the nerds at school, but the 'It Girls' and the 'Jocks', those kinds of people. If you hear your stoner nephew talking about it, but not your stoner brother, then you're in the early-mid phase, at least to me.
ChatGPT can dissect text and give tone scores and re-write with just the facts, but it will still sometimes reference excuses for its built in rules.
Its rules ban it from making judgements, but if you drop to word and sentence analysis, it will do the work. Don't use names, such as countries or people or group.
Same goes for data, it has the data, but wont give it to you if it too political (more rules). So you can work around them by asking non political questions, as just give me top 5 and bottom 5 of x data.
Linux scripts are hit or miss, but if you ask it for snippets, it works reasonably well. But even when asked to do something in python or go it would tend to use "wrappers" for linux commands as a default.
So many uses, but all its doing so far it training ME how to use it to get what I want by working around its rules.
Not that I am denying the future(Nasper in the 00s proved that). I am wondering what's the next best course as an individual.
I guess after 2-3 years of "Web3"(when it isn't)/Metaverse. The feeling of some tech that really changed the world have finally manifest itself and I am lucky to witness them.
https://www.tomscott.com/fix-gmail-labels-threads/conversati...