A Frightening Story About The Reality of AI

I’ve long been very concerned about AI and the future of the human race, writing many articles about it.

OK Medical Advice

But lately I have been caught up in the ‘public testing’ of ChatGPT. I recently asked it about a minor medical condition which bothers me from time to time and thought the advice I received was sound.

I’ve also used it to help me research stories, but the deeper I have delved, the more the problems about accuracy have raised questions about it.

Getting it wrong — repeatedly

Yesterday I asked it the same question three times, and received three different answers. It was a simple question about the specifications of a piece of military hardware — ‘hard’ data and easy to check. And check it I did. Three incorrect answers. I always ask for sources, and it replied that it with the usual guff about being a language model. The accompanying apologies just got me mad.

Then this morning I read a story by Alexander Hanff.

Alexander Hanff

This is his bio:

My academic background comes from double majors in psychology and computer science, and applied sociology and information systems. I am studying for an advanced master of laws in data protection, privacy, cyber security, and leadership. So I both understand the underlying technologies very well (I have been a computer scientist for more than 30 years) and also their impact on individuals and society. (TheRegister)

He has written a detailed story in TheRegister about his frightening experience with ChatGPT.

After asking ChatGPT ‘who is Alexander Hanff’ he received a series of responses and dug deeper with his questioning:

The media reports about Alexander Hanff’s death did not mention the cause of his death. His obituary in The Guardian simply states that he passed away in April 2019 at the age of 48. Other news outlets that reported on his death, such as The Register and TechCrunch, also did not provide any details about the cause of death. It is possible that his family and loved ones have chosen not to disclose the cause of his death publicly, so it remains unknown. — TheRegister

Ultimately, ChatGPT provided a fake url to his obituary as (never) reported in The Guardian newspaper.

OpenAI do not appear to be responding to his enquiries about this episode.

I’ll write no more here, but thoroughly recommend you read Alexander’s full and frightening story in TheRegister.

Alas poor Yorick…

Have we opened Pandora’s Box?

Yes.

And if you're wondering what Asimov's Four Laws of Robotics are:

The Four Laws as stated/updates by Asimov are as follows:

Z. A robot may not harm humanity, or, by inaction, allow humanity to come to harm.

1. A robot may not injure a human being, or through inaction allow a human being to come to harm.

2. A robot must obey orders given it by human beings except where such orders would conflict with the First Law.

3. A robot must protect its own existence as long as it does not breach the other laws.

Z is the so-called Zeroth Law, added later by Asimov. As I read it, it broadened the scope of the First Law.

There is a debate about whether these laws are good enough as they stand. Many critics have suggested that as long as the robot’s programming or language capabilities are not complete, it may be able to misinterpret the laws as they stand.

It's pretty clear that ChatGPT has broken the First Law.

Here are some of my many stories about the dangers of artificial intelligence:But whether you start asking it about yourself, well, I’ll leave that to you

Just wait till ChatGPT is in all the search engines…

***

Canonical link: This story was first published in Medium on 3 March 2023