MacMAD Meeting Topic, Tuesday, March 19, 2024
Artificial Intelligence (AI) has meant different things at different times to different people. It is a moving target. In the 1970s, it meant a machine playing chess. Today, no one would call that artificial intelligence. Now it might be a machine generating an essay, drawing or photograph from a short prompt.
Artificial Intelligence has become vastly more powerful, and it will only continue to get better.
Whether you realize it or not, you are already using Artificial Intelligence in several forms:
- Facial Recognition – “Face ID”, Photos
- Recognition of Objects: cat, dog, building, The Eiffel Tower, vehicles, humans
- Voice and Speech Recognition, e.g. Siri commands
- Siri Suggestions – AI forecasts what App/Destination/Search you might want now
- Sound Recognition, e.g. Siri can recognize smoke detector sounds
- Navigation Apps, driver assist, “self-driving”
- Language Translation, Apple & Google Translate Apps
- Speech Generation – Personal Voice, Live Speech, in Accessibility Settings
- Recognition of facial expressions and gestures – Used in Face Time and the Vision Pro
Face ID is an example of very unobtrusive AI, probably the best kind. You are usually unaware of Face ID unlocking your phone except in the rare situation where your face is obstructed for some reason and it doesn’t unlock.
Apple has dedicated about 1/4 of the on-chip real estate in their new processors to the Neural Engine. This shows a commitment to AI going forward at Apple. The Neural Engine does on-device machine learning.
All Your Internet are Belong to AI1
Much of what you see when you browse the internet is determined by AI.
- Search Engine Rankings
- YouTube Recommendation Engine
- Amazon “Inspired by your Shopping Trends”
- Video recommendations for Apple TV, Netflix and Amazon Prime
When you do those captchas (“I am not a robot”, “click all images containing a bus”), you are providing training data for AI algorithms.
Generative AI
The new type of AI getting a lot of attention in 2024 is Generative AI, so called because it generates something, whether text or an image.
When you hear about ChatGPT, GPT stands for Generative Pre-trained Transformer. This is a large language model trained on a large amount of text.
This type of AI tool can be used to generate documents like reports, emails, ad copy, blog posts and computer code. The generated text usually looks very good, but you must check it for accuracy.
How Can I Try Out These AI Tools?
It’s early days in this phase of the AI revolution, and things are changing rapidly. Web sites appear and disappear. Companies are bought and sold. Tools change hands and change names. Here is a list of generative AI tools that work as of March 2024.
- Claude.ai – Account required. Free and paid tiers.
- The Microsoft Copilot App – May require Microsoft 365 subscription? Includes Designer (for images), Vacation Planner, Cooking Assistant and Fitness Trainer.
- Microsoft Bing – Copilot web interface
- Google Gemini – Gemini web interface
- Typing Mind – accessible via Setapp
- Photos Revive App – Photo colorization – accessible via Setapp
- MacGPT – Mac App interface to ChatGPT.
Apple is reportedly negotiating with Google and/or OpenAI to incorporate some of their Generative AI tools into the iPhone.
Also worth a look:
Sora – from OpenAI, generates videos from prompts – only pre-recorded demos now
Artificial General Intelligence
Artificial General Intelligence (AGI) doesn’t exist yet. Existing AI software is powerful in certain domains, but it is helpless outside those domains. ChatGPT can write a passable paragraph for a PhD thesis, but it’s not as smart about the world as a five year old. True AGI would have intelligence equal to or better than human intelligence in at least all the domains where humans are knowledgeable.
Here’s Copilot Designer’s response to a request for three yellow ducks and a green goose. It totally failed to include a green goose, or any green bird. It’s really not too bright yet. It draws better than a five year old, but lacks understanding.
One theory is that in order to develop AGI, AI will have to be embodied in a robot where it can gain the experience of the world that a child has.
Is AI a Threat to Humanity?
There have been a number of topical headlines recently suggesting that AI is a serious threat to the human race. Caution is warranted, as with any new technology, but I believe that the threat isn’t Artificial Intelligence itself, but our own lack of wisdom in deploying it.
Most of our fear of alien or artificial intelligences is that they will be too much like us. We know all too well the human capacity for fear, greed and desire for power.
Artificial Intelligence is likely to lack those things unless we are stupid enough to design them in. Without an equivalent to the more primitive parts of the human brain, robots or AI will lack lust, greed and even an instinct for self-preservation.2
We obviously shouldn’t give an AI control of nuclear weapons, but that has more to do with the danger of such weapons than of AI.
A more realistic concern in the near term is that AI may threaten many people’s jobs. If you write ad copy or blog posts for a living, AI is already coming for your job. If you are an illustrator or photographer, AI can already accomplish many of your tasks in a fraction of the time. Lawyers and computer programmers are probably next, and I am sure there are many other job categories at risk.
Like any other tool, AI can be used by bad people to do bad things. Since it is a powerful tool, it can have a big effect. It could be a propaganda tool without precedent, and will no doubt be used to spread misinformation, fear, uncertainty and doubt on a massive scale.
These are serious social problems. We have already experienced social problems from other aspects of technology such as automation and social media. As a society we need to figure out how to fairly distribute the wealth and other benefits derived from AI and similar technologies and how to protect society from misuse of AI.