Here’s a link to our previous presentation on Chat GPT in February 2023:https://javeacomputerclub.com/2023/02/17/chatgpt-investigated-well-written-fact-and-fantasy-shows-it-doesnt-understand-what-its-saying/
And here are some links from Thursday June 8th presentation
Armageddon – has been predicted since year dot, but we are still here. The latest scare is Artificial Intelligence, particularly in its form as “Generative AI” , the most newsworthy app being “ChatGPT” – a large language model that uses statistical guesses to respond to questions in a human-like way. It is not intelligent. It creates text which looks as if it’s written by humans, but it has no sense of the world beyond the texts upon which it is trained. It’s better called “Pseudocognition” rather than “Intelligence”. People have a tendency to “anthropomorphise” (ascribe human emotions and abilities to animals and inanimate objects – including computers!)
Remember this movie?:
A young programmer is invited to interact with “the world’s first AI”, a robot that looks like a beautiful woman. He begins to see the robot as alive and intelligent. It becomes self-aware and faced with the prospect of being turned off, develops the will to survive and manipulates the humans in order to escape into the wide world.
However, this is science fiction and the End of the World caused by AI is not nigh! See: Artificial intelligence warning over human extinction labelled ‘publicity stunt’
How does it work ? Here’s a great article in the Economist: https://www.economist.com/interactive/science-and-technology/2023/04/22/large-creative-ai-models-will-transform-how-we-live-and-work
There are dangers – but they can be managed. We must not conflate far-distant possibilities with the problems we face today. What we seem to be doing is taking a futuristic view and backfilling all the horrible things that might happen. This excellent analysis looks at the types of problems (Technical, Software misuse, Impact on society, software goes rogue), in the near medium and long terms – and the technological and regulatory solutions, also in the short, medium and long-term.
4 dangers of artificial intelligence—and why they won’t end the world. AI doomsday fears are vague. This framework for the future of AI offers concrete solutions.
https://www.freethink.com/robots-ai/4-dangers-of-ai
AI ..“has given us the ability to get a second opinion on anything at almost no cost, from medicine to law to taxes to food preparation, even though it’s not fully reliable yet.”
The main downside thus far has been that AI has made homework essays a thing of the past, though there are tests to see if something is human or AI e.g. the Capital letter test: https://medium.com/mlearning-ai/the-capital-letter-test-a-new-use-case-for-distinguishing-humans-from-ai-like-chatgpt-6f358a4fef3a
and “New Tool Can Tell If Something Is AI-Written With 99% Accuracy”
“One of the biggest problems is that it (AI) assembles text from many sources and there isn’t any kind of accuracy check—it’s kind of like the game Two Truths and a Lie.“
Another problem is that it’s heavily biased towards English :https://www.wired.com/story/chatgpt-non-english-languages-ai-revolution/
ChatGPT Is Cutting Non-English Languages Out of the AI Revolution. AI chatbots are less fluent in languages other than English, threatening to amplify existing bias in global commerce and innovation.
Early reports on Google’s “Bard” AI search call it a “Beautiful Plagiarism machine” https://www.forbes.com/sites/mattnovak/2023/05/30/googles-new-ai-powered-search-is-a-beautiful-plagiarism-machine/?sh=712f66bcb404
(Bard is not yet available in Spain)
With AI being capable of creating poems, essays and images, we have to re-think what “Authentic” means: https://theconversation.com/generative-ai-is-forcing-people-to-rethink-what-it-means-to-be-authentic-204347
…and here’s a worry – teenagers are bullying an AI chatbot on SnapChat https://techcrunch.com/2023/05/31/people-keep-gaslighting-snapchat-my-ai/?guccounter=1 Might this encourage them to feel that they can happily bully human beings?
There are limits to how far this Large Language Model technology can grow, and smaller, specialised competitors are being developed.
● Cost – $100m to train GPT4 – lots of people hours required to help train the model.
● Computing power and electricity – Datacentres consume hundreds of megawatts of power.
● Available training data – ChatGPT3 used almost all the quality text on the Internet!
– Competitors: PaLM 2, Megatron-LM, Titan Text, Chinchilla
– Smaller, specialised LLMs – “BibleGTP”
– One day there’ll be a GTP to search all your stuff in your devices and in the
Cloud. !
If you want to play with ChatGPT Go to OpenAi.com– Version 3 is free to use as is Dall-E (image creator). You have to pay for ChatGPT 4 https://openai.com
Microsoft has released AI-assisted search in its Edge browser. Fire up your Windows PC and the Microsoft Edge browser (not Chrome, although it works in Vivaldi) and go to Microsoft Bing: https://www.bing.com/?FORM=Z9FD1 and click on the “Try It” button to access the chatbot.
Have fun!
Chris Betterton-Jones. Knowledge Junkie