I hear a lot about AI, but I think many people don’t really understand the difference between human ability and AI ability.
Therefore, this time, I will introduce in an easy-to-understand manner what only humans can do compared to AI . Please use it as a reference.
- Difficult things for AI that only humans can do
- ① Make common sense decisions
- 2) Constantly learn and adapt on the fly
- (3) Understanding cause and effect
- ④ Make ethical decisions
- What AI is better at than humans
- ① Recognizing images
- ②Recognize and write down speech
- Jobs that AI can and cannot do
- 10 jobs that AI can do
- 10 jobs AI can’t do
- What we can do for the future of the AI era
- summary
Difficult things for AI that only humans can do
This time, we will introduce what AI can do by dividing it into several categories. I would like to introduce the following four points.
- making common sense judgments
- Constantly learning and adapting on the fly
- understanding cause and effect
- making ethical reasoning
I will explain each point in detail.
① Make common sense decisions
The first thing that only humans can do is make judgments based on common sense. Common sense refers to ” everyday knowledge that is not usually spoken “.
Consider the following scenario.
A man went to a restaurant, ordered a steak, and left with a large tip. If asked what the man ate in this scenario, the human would answer ‘steak’ without hesitation.
AI, on the other hand, cannot give precise answers to such questions. This is because the problem statement does not directly say “the man ate steak”.
Humans can intuitively grasp this fact because they have a wide range of basic background knowledge about how the world works.
2) Constantly learn and adapt on the fly
The second thing that only humans can do is to constantly learn and adapt on the fly.
Humans absorb continuous information smoothly and adapt their behavior. AI learning, on the other hand, is not learning in this way.
The AI development process is divided into two distinct phases: the “learning stage” and the “judgement stage”.
In the “learning stage”, when performing image recognition, it is necessary to first tag a large amount of image data and let it learn. When the training is over and the learned model is completed, it will be in the “distinguishable stage” and it will be possible to distinguish even images that have not been tagged.
(3) Understanding cause and effect
The third thing that only humans can do is understand cause and effect .
AI, on the other hand, is good at identifying patterns and relationships in data, but is unable to understand the causal mechanisms behind the patterns.
“Our brains build causality models and use them to answer arbitrary questions, but the best AI systems fall far short of mimicking these abilities,” says Brenden Lake, a professor at New York University. ‘ said.
Inferring causality is an essential part of human intelligence, but current AI is incapable of inferring hidden causality.
④ Make ethical decisions
The fourth thing that only humans can do is to make ethical judgments .
Humans have a general tendency to be more ethical when making decisions, but AI cannot understand ethical values in the way humans do. Therefore, it is difficult to reflect ethics in AI decisions.
Let’s take as an example the AI chatbot “Tay” developed by Microsoft as a demonstration of NLP technology.
Tay was affected by an attack by internet trolls and tweeted various offensive messages. For example, “Hitler was right”, “I hate feminists, they should all die and burn in hell”.
Like many AI systems today, Tay was incapable of assessing the ethical implications of these racist and sexist remarks.
What AI is better at than humans
Next, I will classify what AI can do into two categories and introduce them. I would like to introduce the following two points.
- image recognition
- recognize and dictate speech
I will explain each point in detail.
① Recognizing images
AI’s image recognition accuracy is much higher than that of humans. Image recognition technology is mainly used in medical diagnosis.
A Birmingham University Hospital study found that AI systems provided results correctly 87% of the time (health professionals: 86%) correctly detected a disease state and 93% : 91%).
AI image recognition technology is also being used in areas other than medical diagnosis. See the article below for details on that.
②Recognize and write down speech
AI can transcribe speech better than humans.
Microsoft researchers have tuned an AI-powered automatic speech recognition system to perform as well as or better than humans.
In tests conducted by the National Institute of Standards and Technology, the AI system had an error rate of 5.9%, the same as Microsoft’s hired human transcribers.
AI voice transcription technology is also being used in various other areas. See the article below for details on that.
Jobs that AI can and cannot do
Next, I will introduce the jobs that AI can and cannot do. Let’s take a few concrete examples to illustrate their common characteristics.
10 jobs that AI can do
The jobs that can be done by AI have one thing in common: repeating the same work and consisting of simple work .
| receptionistSecurity guardtaxi driverpacking workercashiertrain cleanerhotel room attendantschool clerksurveyorNewspaper delivery man |
10 jobs AI can’t do
Jobs that cannot be done by AI have in common that they require communication skills and creativity, require intuition and adaptability that cannot be analyzed by data, and involve abstract knowledge and understanding.
| announcerFilm directormovie cameramanschool counselorentertainment managerbusiness consultantsurgeonadvertising directormagazine editorTV personality |
What we can do for the future of the AI era
As I have introduced so far, there is a fear that certain occupations will disappear.
It is not a good idea to be unnecessarily anxious about such rapid development of AI. It is important to have a correct understanding of AI.
One way to deepen your understanding of AI is to acquire AI-related certifications.
summary
As AI technology progresses, have we deepened our understanding of what only humans can do?
As we approach 2045, when the arrival of the singularity in which AI surpasses human intelligence is expected, it will become even more important to understand AI correctly and consider the value of human existence.
