Current AI is totally dependent on a) the data it is fed b) The algorithms and programming that tell it how to use the data.
1) AI data in incomplete. The areas of incompleteness are not uniform. Incomplete data is clustered in certain areas. This makes AI biased.
2) AI cannot tell the difference between fact and fiction, fact and opinion. Example: It treats data from SPLC as fact when it is opinion. It treats Cabot Cove ME and Spart MS as geographic fact when it is fiction.
3) AI is intentionally programmed to not be racist. Thus, it cannot relate factual data about race in crime statistics, effectiveness of attempts to fight racism.
4) AI algorithms and logic to process data is full of arbitrary guidelines. Much of this is for good reason and the difference between NARROW and GENERAL AI. NARROW AI is focused on driving a driverless car, or finding the cause of an illness, or other narrow task.
At the edge of the narrow scope AI makes many errors.
But General AI makes even more errors due to its lack of expertise in a specialty.
Current AI can be improved. There are 5 or 6 brands of AI. If one AI does better at accuracy and a 2d does better at speed and a 3d does better at cost efficiency which is going to get the market share of the business and money?
The company that buys the other two and merges the technology.
The one that can leverage the others to its advantage.
AI is programmed to lie about big tech. It has pre-programmed word salads about:
1) H1B tech related worker immigration
2) Censorship
3) Tech Monopolies
It is a waste of time to bring up these topics.
AI might be accurate, fast, cheap.
You can have 2 of the three.