AI - Fear leads to the dark side #20

5 AI questions answer by a human, with a little help from AI

Hey all, today we're in for a special treat - I'm putting on my AI hat and answering 5 questions/thoughts related to AI.

Within the newsletter, you’ll find links to 2 different AI systems you can try yourself.

The list comes from comments made on a TED post on Facebook.

The purpose of responding to these questions/thoughts is to educate people about AI.

Before we start, to recap, I have spent the last 10 years studying and working with Machine Learning and AI field, and have completed an MSc and PhD in the subject.

Also note, these questions are not technical questions and come from real people. The challenge with AI is that the average person is only exposed to it through film.

^ this image was generated by AI.

I don’t have a good feeling about AI at all, and I wish we could put the brakes on it. Just because we can do it doesn’t mean we should. I have no confidence that adequate safeguards and restraints will be applied to it. I believe it will be used to manipulate people in all sorts of ways.

This comment seems to be generalised fear. You could literally replace the word 'AI' with most things. Give it a try...

I don’t have a good feeling about <TV/Radio/Cars/Internet/Fire> at all, and I wish we could put the brakes on it. [...] I believe it will be used to manipulate people in all sorts of ways.

I’m fairly confident that someone, at some point in time would have said something similar about many things.

We can be fearful of most things and also a lot of things can be used to manipulate people.

There is an AI system/approach out there called Deepfakes. It is a system which generates videos of real people saying things. These videos can show a person who looks and sounds like a real person saying anything you want. Obviously, it could be used to manipulate people.

Even Elon Musk has been targeted with Tesla stating “[Musk], like many public figures, is the subject of many ‘deepfake’ videos and audio recordings that purport to show him saying and doing things he never actually said or did”.

Perhaps a little off track, but the extreme example of a deepfake would be seen in Futurama with the Lucy Liu bot. But that’s an extreme example and not real, it’s just a cartoon.

But in the statement above, let's remove AI and replace it with Twitter...

I don’t have a good feeling about <Twitter> at all, and I wish we could put the brakes on it. [...] I believe it will be used to manipulate people in all sorts of ways.

According to an inquiry, Donald Trump's 'Wild' tweets incited an attack on Capitol Hill. A tweet or series of tweets from a real person was able to 'manipulate people in all sorts of ways.'

I guess the point is, people are already been manipulated.

As a writer by trade, I view it as an existential threat. Then, I remember I have faith that there is some measure of humanity in art that machines are incapable of creating or replicating.

This is kind of challenging to respond to due to the subjective nature of 'art'. We all can look at art and see/feel different things.

Also, my interpretation of this question is both text and art (for ease - a picture/photo/painting)

But let's look at art - yeah, AI can produce artwork.

Based on Alan Partridge`s TV show idea, I used Bing Image Create to create the following images simply by typing "monkey tennis".

Pretty nice eh? They are all AI generated pictures.

Or what about asking ChatGPT to write me a haiku about fearing AI.

Machines of the mind

Unknown power lurks within

Fearful of their might

I can’t write a haiku.

So yeah, AI can do these things but, and here is the huge but...

They are all trained on work that humans created. I think it was ChatGPT v3 which was trained on 570GB of data from books, web texts, Wikipedia, articles and other pieces of writing on the internet. In total, 300 billion words were fed into the system.

And it looks like DALL-E, which is the underlying AI system used to create the monkey tennis pictures, was trained on 400 million pairs of images with text captions scraped from the Internet.

For me, this results in 2 more prominent questions:

• Ownership/Copyright - are the AI systems trained on data they are allowed to use? Microsoft, GitHub, and OpenAI have already had a court case brought against them for this.

• Can we ever replace the human element in the original? Current systems are trained on human-created work to derive their own work. By only training using AI-generated images or text, it could essentially be the equivalent of inbreeding, where subsequent generations result in deficiencies and abnormalities.

Also, I missed commenting on the humanity in art which machines are incapable of creating or replicating. Here is a picture of giant turd sculptures by the Viennese art collective Gelatin (created by humans).

I asked AI to make me a picture of a ‘turd in a museum’, but it declined to do so… IDK who wins in this situation.

I have doubts on whether it will be used to create a more compassionate and humane world.

Yep, it probably won't create a more compassionate and humane world.

After all, humans already have the ability to create a more compassionate and humane world - but they don't.

However, AI is used to predict "effective drug combinations to fight complex diseases faster" and also to "predict gene expression from DNA sequence". I don't know much about gene expression, but I assume it’s important.

So it might not help create a more compassionate and humane world, but it is helping us advance science and medicine.

Before it is widely used, we need a generation of adults with critical thinking skills who can look at a photo, article or video and look at what is being presented with an analytical eye.

Don't even get me started on this one. We need this without even talking about AI.

How many people do you know who read one single thing and believe it fully, and then we have the news/media propagating fear.

I have even heard people say “it must be true, it was in the news”. Oh, dear.

Did you ever hear that the EU was banning Bendy bananas? Well the British people did, and it featured in the news.

Lots of people read the article, lots of people believed it to be true, and lots of people got very angry.

Not that many people applied their critical thinking skills, then simply laughed at it and thought it was probably not true.

The EU even had to keep track of all the misinformation that was spread by the media. And I don’t know what happened to the missing £350 million per week the NHS was promised.

So yes, we do need people to develop critical thinking skills, 100% - but that is nothing to do with AI.

Also, I asked AI two questions...

I asked: “Make up a bullshit newspaper headline about bananas”

The reply: "Banana shortage leads to nationwide banana bread baking frenzy"

Next question I asked: “Make a newspaper headline about bananas which will make people scared”

The reply: "New deadly virus found in bananas spreads rapidly across the globe"

It did a pretty good job.

So good in fact that I think AI could put some national newspaper journalists out of a job.

I am afraid AI will serve those who made it, the big tech giants, rather than humanity.

Yep, likely, but with a big BUT.

And taking of Big Butts, did you know the porn industry drives a lot of tech innovation? Not sure how much they are using AI or deepfakes but you never know - I guess we’re back to the Lucy Liu bot thing again.

Butt, back to the point (see what I did there) - Big Tech is driving a lot of AI innovation.

Deepmind was one of the pioneers and was acquired by Google for more than $500 million. OpenAI, which created ChatGPT, saw Microsoft invest $10 billion earlier this year. And Meta has its own team called Facebook AI Research (FAIR).

The great thing about FAIR is that they open-source their research. They even recently opened up their LLaMA model (equivalent to ChatGPT) (https://ai.facebook.com/blog/large-language-model-llama-meta-ai/)

In addition, they have the open-source Pytorch framework which is used to build AI systems.

Let's be honest, where is big tech using AI? My guess is 2 places:

• trying to recommend content for you to keep you hooked on the platform (TikTok are a great example of this)

• then, trying to give you the optimal advert to maximise the amount of money they and their advertisers make.

AI should not exist. Mankind has its own intelligence. Not using our intellect or learning is an insult to our very existence.

OK, so let's answer this question with a history lesson.

Want to know where this all started? Alan Turing is widely considered to be the father of theoretical computer science and artificial intelligence.

During World War II he created the device which cracked the ‘Enigma’ code. So yes, mankind does have its own intelligence, but that intelligence sometimes isn't enough. In this example, they needed to create a machine to solve the problem.

AI systems can be simplified to...

Finding the complicated relationships between the inputs (things you put into the system) and the outputs (the things you want out). The more complicated the relationships are, the more sophisticated the AI model will be.

Handling text and images are complex problems.

Finally...

I asked an AI system "should we be frightened of AI", it’s response…

As an AI language model, I cannot have a personal opinion, but I can provide information based on data and research.

AI has the potential to bring about significant benefits in various fields, including healthcare, education, and transportation, among others. However, like any technology, it also has the potential to be misused and cause harm.

The fear of AI is often rooted in science fiction, which portrays AI as a threat to humanity. While AI does pose some risks, researchers and policymakers are working to develop safety measures and ethical guidelines to mitigate these risks.

Ultimately, whether or not we should be frightened of AI depends on how it is developed and used. As with any technology, it is essential to proceed with caution, anticipate potential risks, and prioritise safety and ethical considerations.

Fin.

I hope you found this useful.

As always, wishing you a great weekend.

Best wishes, John