Despite all this, Meta is keeping “Blenderbot3” out there in public… Why - “Meta has made the BlenderBot 3 public, and risked bad publicity, for a reason. It needs data.”
Things such as this just underscore how hard, hard AI is, and how far we are from any sort of general-purpose artificial intelligence, let alone one which is actually self-aware.
As regards the end-result of all this data, more folks should read up on the history of telephony in the Soviet Union. If Lenin had had access to modern computers, his successors would have not had any interruptions in their control.
Very true. I’ve played with some of the very specific or task focused AI or AI-type interfaces and even they struggle with some of the tasks. Being from the world of art & design, I’ve played with Dall E and the thispersondoesnotexist service powered by Nvidia.
I have tried Dall E which I found quite basic but interesting (I believe the new Dall E 2 is quite powerful - but I haven’t had the chance to try it) and thispersondoesnotexist tends to produce quite similar facial structures - there’s definitely a “look” with thispersonetcetc
Anyhow - they are very specifically focused whereas I can’t imagine just how much data and processing power would be required to run a general chatbot AI for social networking
Ages ago, there was a utility program, “WordWeb” which ostensibly grew out of a semantic list of connected words from Artificial Intelligence research — I installed it a couple of times, but never really used it.
It’s thirty years away. It has always been thirty years away since the 1950s. e.g. Arthur C. Clark’s 2001 HAL 9000 was thought by tech people in the late 1960s to be a reasonable prediction. :vb-agree:
“Within thirty years, we will have the technological
means to create superhuman intelligence. Shortly after,
the human era will be ended.” - from the 1993 abstract.
That’s next year!
(Vinge revised the date to the mid-2030s in the 2008 IEEE Special Report on the Singularity symposium, sadly now behind a paywall. The collection of articles by experts, pro and con, was entertaining and informative for its time.)
Useful to know, I’ll adjust my survival kit now… it’s already larger than I can carry - so far I have
Nuclear war survival kit
Global Warming survival kit
Stockmarket crash / 2nd great depression survival kit
and now Super human AI survival kit to add to it.
Good thing I don’t live in the USA as I’d need to add a “New World Order” survival kit too.
More seriously, I think AI is always going to be 30+ years away until we decide exactly what we want it to do and what parameters it would work within. A general knowledge / General purpose AI like you see in movies answering all sorts of questions and running all sorts of tasks and routines (such as the “computer” in Star Trek or superAi that runs the Matrix) and thinking for itself would (I believe) require resources and hardware beyond anything we can create right now or in the next 50 years.
On the other hand - task specific AI like a HAL 9000 companion on a space flight is, I think, a lot nearer to realisation.
As long as Elon Musk doesn’t go and call it HAL 9000…
Fo those that might want to dive a bit deeper in to some informed AI future speculation. This book is a good place to start. It’s a collection of short stories but informed by both the current state of it as well as what informed thinkers, believe what the future might bring.