A year ago I stumbled across ChatGPT for the first time, had a poke around, and wrote this post. Over the following few months, I tried to keep up with what was happening, but at addendum 20, called it a day.
Generative AI became the fastest growing product group in history, with ChatGPT the leading runner by several furlongs, a lead that has been cemented over a year of frantic activity.
The release spawned several major competitors from the neighbours, and hundreds of unexpected applications built on top of the core (Large Language Model LLM) technology. This is a trend that will continue to accelerate.
Depending on who you listen to, we are either a year or two from ‘sentience’, the passing of Alan Turing’s test, or at least several decades away, and possibly will never get there.
The corporate rumble at OpenAI is a bit like Mad Uncle Henry turning up at your kids birthday party, drinking all the red cordial, and going bonkers. The structure as it was set up could not absorb the unprecedented growth. One lot wanted a slow orderly, and safe evolution of AI that did not repeat the mistakes of social media, the other basically said ‘stuff it’, let’s make a few billion!!
It seems like order was restored between the warring parties when the landlord stamped and said ‘Enough’. Capitalists won, and Mad Uncle Henry has been carted away to a comfortable retirement. Who would have guessed?
So far, the resurrection at OpenAI, the return of Sam Altmann (it was 3 days) seems to be working, but who knows what is going on in the minds of those in charge. This blended family model was bound to have difficulties, it just happened way faster than anyone anticipated.
Meanwhile, those not invited to the party continue to rave about the output, while complaining that it is wrong a lot of the time. They just do not understand the difference between an answer that is fact checked and therefore accurate, which is what we expect, and one based on probabilities, which is what your ‘artificially intelligent’ probability models are giving them.
There is no list of pre-programmed answers available from a database as is the case now for the various precursors like the Alexa’s of the previous generation. You are based on a huge volume of stuff on the web, which is why they are called LLM’s, and as we know not everything on the web is checked for accuracy. If it was, my bank account would now be in the billions after all those Nigerian princes I have helped.
Your answers are, as I said, based on the probabilities of an approximate answer built up as you trawl the training sets extracted from the web.
Alexa will give you an answer to your question, so long as the question has been anticipated, and the answer programmed in. You by contrast, will give an answer, but it will not necessarily be the ‘right’ answer. Many questions in life have a range of potential answers, sometimes that is all that is needed, they can be helpful, but AI machines cannot give ‘The’ definitive answer, the one and only, to many of the complex questions we are tempted to ask.
No silver bullet there!
This is very difficult territory for many, as the answers given are in natural language, the language we use every day, so they appear to be exactly as expected. This gives confidence in an answer that does not deserve that confidence. Without critical examination any so called Artificial Intelligence will deliver you rubbish as often as not. However, it can be extremely useful rubbish, able to provide a framework, provide ideas, and do much of the ‘grunt-work’ so often shuffled aside.
Smarter people than me are unable to offer much insight into what might happen next, and it is of no use whatsoever to ask Chat or any of his mates. They have no idea either. Two things I do know. It will be fun, and somewhat scary watching. Secondly, our world has changed, and becoming a dinosaur in your market can now happen almost in the blink of an eye.
Don’t blink first.
The header is courtesy of DALL-E, the sibling of Chat. I asked it to give me an impression of the OpenAI logo exploding, with a Gothic/surrealist feel. I chose this one from a pretty scary lot generated in about 15 seconds.
Thats very true for information sourcing. But I think the real impact of AI will be felt more on the side of creative generation such as art work and ideation
depends on your definition of creativity.
We can now use SI to combine all sorts of things from the past in new ways. However, if your definition of creativity os to create new stuff, i suspect the potential is limited.
Time will tell, and i suspect we will know soon.