According to pandata.co 3 ways AI can fail
1. Racism can be in AI responses to questions
2. Bias against women (or other groups) can creep into AI responses.
3. AI deep fakes pretend to be CEO to help out criminals or other activists to further political agendas, at the very least creating an atmosphere of disinformation.
5 times AI failed us according to jumpstartmag.com :
1. Tesla crashed due to auto pilot failure (also driver inattention, but the driver assumed safety when there was little)
2. Amazon’s recruiting AI showed bias against women.
3. AI camera mistook a bald head for the ball (as it was tasked to follow the ball), so viewers saw the linesman a lot.
4. Microsoft’s AI chatbot Tay has racist and sexist tendencies.
5. False facial recognition turns into a black man’s arrest spending 10 days in jail. Later the case was dismissed.
Unfortunately several issues arose where the facial recognition software pointed to the wrong person in New Jersey.
10 AI failures listed in a Linked post by Darpan Kulshreshtha:
- Microsoft’s Tay racism and sexism explained( including how quickly it went bad)
- French chatbot test suggested suicide when asked a question.
- Minority report
- Are smart underwear really smart?
- Tesla car crash
- Bald head issue in ball game
- Amazon’s recruiting tool went bad
- Google Ad bias
- Robot jail break.
What about the latest interview with Elon Musk??
Mr. Musk says that AI has the potential to create more disinformation than we have seen so far.
It is my point with the image I had created:
I wanted to show that AI has the potential to become King or an authoritarian with a mustache if the ethics are unbound. One must create specific ethical constraints – and continue to test the system for one to be comfortable that AI will not become something we would not wish on our worst enemy.
Elon Musk interview about AI is creating waves now (youtube video)
The key is to make the AI with ethical intent otherwise many issues can arise as you can see above (sexism,racism and other problems). having an unbounded AI we are finding to be unpalatable.
Remember AI is in many programs being built in now – such as Netflix, and TikTok, as YouTube tries to give you more videos to watch. Are the videos that you should continue watching (with more ad dollar$) a good AI or bad AI? We should expect some transparency in our AI so that we can tell what is going on.
This is why I like You.com as an AI app instead of chatgpt since it gives the link where it found the info to your question.
Contact us to discuss.