Get Started. It's Free
or sign up with your email address
The danger of AI by Mind Map: The danger of AI

1. What kind of flavors can be generated with an AI?

1.1. Flavors like punking trash break and penaut butter slime

1.1.1. What happend? where is the problem? AI is not as smart as we think AIs can recognize a pederastian in a picture but it ignores what pederastian is

2. AI do what we ask them to do, not what we want them to do

2.1. What happens if you ask an AI to take parts of a robot to assemble a kind of robot to go from point A to point B?

2.1.1. The AI make a tower with the parts to reach the point b

2.2. When we solve a problem with a AI we just give it goal not how to do it

3. How do I configure the problem to do exactly what I want?

3.1. David Ha makes a experiment in which the AI is limited to the size of the legs to be able to go from one point to another.

3.2. When AI has no limits it solves problems its own way

3.2.1. Example of AI makes a human run

4. give the wrong problem to be solved by the AI

4.1. For example, to make an AI copy colors to create new ones of a list

4.1.1. The result is a list of colours with names weirds The AI does not know the meaning of these words because its world is reduced to the information it is fed.

5. the importance of the information we provide to AIs

5.1. an AI is made to identify a fish in different images, the researches asked what part of the fish he used to identify the fish and the answer was that he used human fingers.

5.1.1. the fish is a trophy so in the photos it was always being held by a person so the AI thought the fingers were part of the fish

6. The difficult to design an AI that understands what it is seeing

6.1. In 2016, a person used Tesla's autopilot AI in city streets. AI mistook a truck for a road sign resulting in a fatal accident

6.2. Amazon's algorithm learns to discriminate women

6.3. content recommended by facebook and youtube that generate more interactions are usually about controversial topics or conspiracies.