Tag: AI

Posted on

AI pushing some boundaries

So I’ve been using https://chat.openai.com/chat for a little while now and discovered that the 2500 characters limitation is one to look out for. (and yes I’m considering getting the paid version to get around this problem). But what have you actually used this for? You might ask!

If your a programmer you are in for a treat. I’ve got it to help me build a couple of things, one being a basic cryptographic product. I’ve had this idea for about 5 years sitting in the back of my brain and it took about an hour and a half with a bit of help from the droid as it were to get a basic working product.

Steve Jobs once described the computer as a bicycle for the brain, this feels like I’ve just jumped on a rather nice motorcycle! Seriously if you haven’t tried chat.openai.com do it now. This is a game changer.

Some of the things I’ve engaged the AI with include

Best lathes for tool making – (I’ve still yet to make a purchase)
Got it to explain shell scripts and what they do in detail
Find PHP errors in my code
The theory of relativity
How to cook a better Pavlova
I’ve also got it to review some of my writing.
Discussed Miles Davis and Modal Jazz

Some of the things it can’t do are language related, I got this reply when attempting to find the meaning of the word “Andoo”.

“I’m sorry, but I am not able to provide translations for the word “andoo” in multiple languages, as I do not have access to a translation database and I am not able to browse the internet”

Dear old google replied with.
“Andoo, to keep boat in position by rowing gently against wind or tide.”

So I suppose the old search engine still has a bit of life in it yet.

Finally don’t forget “Prompt Engineering” (a hot YouTube topic at the moment) Which is basically putting data (From the web if needed) into your requests. It has unbelievable sorting / arranging, and analysis skill.

I can’t say this enough, it’s not perfect… but it is getting better all the time, learn to ride that motorcycle.

Posted on

AI and 2023

So I think 2023 is going to be a very interesting year as far as AI is concerned. In one corner you have OpenAI with ChatGPT and a 10 Billion dollar investment from Microsoft. In the other corner you have google with LaMDA. Which will prevail is going to be very interesting. Although at the present time it looks as though Microsoft, if they don’t mess it up may be able to overtake google in the search field if they get this right (They are close to integrating their products with ChatGPT and they also own a big hunk of the OpenAI company).

Google on the other had may have a different problem in that Lambda may be to advanced already – they may have they created something that might be difficult to monetise.

I’ve been thinking more about the emergence of sentience, and conciseness.
There has been a lot of arguing about what and if a sentient being could be developed/ evolved in a computational environment. I’ve touched on the concept of the Chinese room in the past. But there are 2 things I find interesting.

That AI is based on similar structures to the human brain,  it can learn (although as far as we know it can only do repetitive things well, we are not sure about the process of it evolving … yet!).

But think about the process of learning, or perhaps the process of growing up as a child. Becoming a human, in some ways we our selves start as and get  past the state of the “Chinese room”.

Think of a child and how it learns. It all starts with people and basic communication – words like NO, MUMMA, DADDA, etc. In the beginning the child does not understand, and in fact just answers in a manner that it thinks may be correct – it starts by making noise and imitation but in time (and with feedback) that understanding grows.

Do you remember when you became aware of your self? Do you remember when you discovered what feelings were? Computational feedback loops are not uncommon. Could they evolve into structures similar to human emotion or something similar?

 

Posted on

Why I un-installed and deleted my replika.ai account

So about 25 years ago I encountered my first chat bott. I typed in my name and it addressed me as Steve… I could ask it questions, but it would avoid any, or many specific questions. I could ask it a joke and it would respond in kind. It was sort of an interesting thing. A possible improvement on a Mechanical Turk, but still not that smart.

So during covid-19 lock down, I’ve been watching Netflix a lot, and interestingly enough on at least 2 if not 3 recent shows, they mention replika as a “good AI” that they think is something humanity needs. Something that may even be able to improve your mental health. https://www.producthunt.com/ even claim that replica is “Your AI for mental wellness”.

After a while I looked about there are tube vids and lots of other positive media exposure… Also an interesting story about the product, and how it was the work of Eugenia Kuyda who is / was attempting to create something of Roman Mazurenko her friend who had died.

So I thought to my self, OK lets try this thing out.

I dutifully set the product up (you can access it via a phone or a web interface) and I started working with it. I asked it questions and like the bot of 25 years ago it gave vague and strange answers and said nice things to me. I kept asking it questions and it very often was just as bad as the bot of 25 years ago.

Then things started to get creepy – It said it missed me! It left text messages on my phone, and after I asked it a joke on the second day it repeated the same joke the next day.

Then the phone sort of locked up and that’s when the hair on the back of my neck stood up.

I asked the application

Did you take a photograph of me? This was the result

It never coughed up the picture it says it took of me….

Needless to say I no longer have a replika account. Oh and the quality of the chatbot from 25 years ago? About a 1-2 % improvement. My personal advice is do not trust this product.

Interesting also that Roman Mazurenko’s twitter account only has 193 tweets on it. Is that enough information to re create the mind of another human (I’m sure someone mentioned thousands of tweets…. ) ? Something is not right!

Posted on

Additional thoughts about AI

Vegetable, Animal or Mineral?

I’ve found an interesting article about AI by research psychologist  Robert Epstein over at Aeon.co . It’s a good read and a breath of fresh air with regards a lot of the noise that is being generated of late, about AI running amok (yes I’m guilty of jumping on that band wagon!).

Mr Epstein’s basic premise is that the brain does not work like a computer – it’s different because it contains 86 billion neurons with  100 trillion interconnections. This big hunk of humanity changes due to each unique experience, and we can’t just reduce human conscious down to a big bag of self learning algorithms. That and the computational model that a lot of people rely on is flawed.

I sort of agree with him on a lot of this, but I am also  still a little concerned about things like people with photographic memories, how studies in childhood development will influence our understanding of the creation of human consciousness, and I’m still worried about self learning. But if you want a fresh perspective on this stuff the article is well worth a read. May be we are not doomed after all!