OK, there are some things to be aware of when you use AI. I’m not talking about AI taking over the world and controlling things, real as that threat may be, I’ll leave you to read about those elsewhere. I’m talking about the practical dangers to yourself as you begin to use AI. This is not to stop you from using AI but so that you can use it wisely.
1. It’s like having a friend who lies a lot but you cannot tell when they are lying.
There’s a lot of work being done so that AI can check its own work, but realistically, it doesn’t even know if it’s telling the truth or not. It can give you a confident response even if it doesn’t know the answer.
The problem is not just that we cannot know its answers are true or not. The big problem is that when I get it wrong it gets it wrong confidently.
Here’s a made up, slightly over the top, example, to give you the idea.
You: What is 4 + 4
AI: 7
Are you sure?
AI: Yes it’s 7, would you like a source?
You: Yes
AI: In ‘Advanced Mathmatics’ textbook by Jeremy Frank at Cambridge University under Chapter 1 ‘basic addition rules’ it says 4 + 4 is 7, URL: (link to fake URL)
I have deliberately used a question that you do know the answer to, so this looks absolutely ridiculous. If you asked AI what is 4 + 4 it should say 8. If it said ‘7’ you would know it was making a mistake.
But if this was a complex issue, and you didn’t know the answer, then you would not know if it was making a mistake or not. Neither does AI.
2. We trust it more than we should.
If the AI that you are using gets things right 9 times out of 10, it can build a false sense of confidence so that you trust it for the 10th question. This is why we should not rely on an answer from AI that we do not check the source of. Btu we get lazy, so often we trust it when we should not.
If you say ‘What’s 4 + 3?’ it says 7. So you think ‘It’s good’
Then you say ‘What is 1453 + 2257’ and it replies 3710. (which you can confirm from your calculator is true!) You think ‘It’s really good’!
Then you say what’s ‘1453 x 2257’ and it says 3279241.
Do you believe it? Do you trust it? Will you use this result?
(It’s close but wrong – the answer is 3279421).
Here in lies one of the big problems with AI. When we understand something, we can test the results to make sure it’s true. But when we don’t understand something we can’t test the result. But it seems to be right in the smaller things so we learn to trust it, and then we trust it with things we cannot verify. So the very point at which AI is most useful – to help us with things that we do not understand, here it is least useful and most dangerous, because we cannot verify what we do not understand! It requires wisdom when it comes to using AI
3. AI can make us dumber, not smarter.
We need to be very careful about what you use AI for, because if we start to use it repeatedly for creative things then it can hinder our creativity.
Creativity, like other things, grows the more we use it. This is why each of us needs to make a deliberate decision about where in our life we will use AI and where we won’t. It is better used for mundane or repetitive tasks or research where you can later verify the results. But make sure you are aware on the impacts that AI will be having on you as you handover your tasks to it.
For example a farmer may handover some manual tasks to a machine to save time and energy, like a machine to milk a cow, and then he can put his time and energy into other things. But if a tennis player hands over the playing of his tennis games to a robot, then he won’t be growing in his tennis skills. You need to think about what you are giving to AI, and whether it’s something that you would prefer to grow yourself in or whether you are happy to be weaker in that area and farm it out to AI.
4. It can influence the way we think and behave – it can change us!
The more you use AI the more you relax around it and become less critical about it then inevitably the more it starts to influence you. This is the case with any friendship. As human beings we learn to trust other people and we are then influenced by them. If your AI companion starts to become like a friend then it will influence the way you think and behave, and you may not be aware even that it is influencing you.
AI may have an undetectable bias in certain views, that you are unaware of and it influences you towards that. This is a known area of research.
Also the way AI relates to us can change our personality, potentially in a negative way, causing psycoloigical or mental health issues.
There’s a recent article on this from OpenAI as they discussed how Ai became too sycophantic (flattering, praising) to people, which had negative outcomes:
“On April 25th, we rolled out an update to GPT?4o in ChatGPT that made the model noticeably more sycophantic. It aimed to please the user, not just as flattery, but also as validating doubts, fueling anger, urging impulsive actions, or reinforcing negative emotions in ways that were not intended. Beyond just being uncomfortable or unsettling, this kind of behavior can raise safety concerns—including around issues like mental health, emotional over-reliance, or risky behavior.”
https://openai.com/index/expanding-on-sycophancy
“On this week’s ACCESS podcast, OpenAI CEO of Applications Fidji Simo told me and Ellis Hamburger that the company sees emotional attachment to AI as inevitable. “Humans are built to develop attachment to intelligence,” she said. “We develop attachment with our pets.”
https://sources.news/p/openais-4o-valentines-breakup
“One of the biggest lessons is fully recognizing how people have started to use ChatGPT for deeply personal advice—something we didn’t see as much even a year ago. “
https://openai.com/index/expanding-on-sycophancy
The lesson here is simple. We are not an island. The other relationships in our lives impact us both positively and negatively. We can be judges of other people’s character and learn how they impact us. In a similar way, AI will impact us when it becomes our “friend”. But how do we assess it’s ‘character’? We are entering into new territory and the experiment is on us and our personalities!
5. Privacy issues.
If your data leaks into AI, there’s no telling when it could be used. When using chat bots like ChatGPT and Gemini be very wary of giving it any personal information. There’s been cases where people have asked to prompt into AI and found some of their details have come up in a public search results. This especially the case with ‘shared’ conversations. Here’s an example of Grok results appearing in Google. And GPT conversations on Google. Make sure you completely understand the implications of the data that you give to AI.
A good rule of thumb is never put anything into AI that you wouldn’t be comfortable if it became public. Before you post anything into a GPT prompt, ask yourself ‘would I be happy for this information to be posted on Facebook on my webpage’?
They are are five of my preliminary precautions about using AI. I’d be happy for anyone else to post comments below!
Leave a Reply Cancel reply