Week 6: AI Safety and Governance

What is the definition of AGI (Artificial General Intelligence)?

• AI systems that can perform specific tasks with high accuracy

• AI systems that can perform any intellectual task that a human can

• AI systems that can learn from experience and adapt to new situations

• AI systems that can communicate with humans using natural language

Advanced AI systems could have which of the following capabilities?

• Superintelligence

• Goal-seeking and resource acquisition

• Lack of human values and common sense

• All of the above

What is the current state of the development of AGI-like systems?

• We already have AGI-like systems that can perform any intellectual task that a human can

• We are close to developing AGI-like systems, but there are still significant technical challenges to overcome

• Developing AGI-like systems is still a distant goal that may not be achievable in the near future

• AGI-like systems are not a priority area of research in the field of AI

What is the timeframe that experts estimate human-level AGI may be achieved?

• 5-10 years

• 10-20 years

• 20-50 years

• 50-100 years

In The case for taking AI seriously as a threat to humanity (Vox 2020), how would AI plausibly harm humanity?

• AIs would eventually become conscious and destroy humanity as an act of defiance.

• AIs could behave in unexpected ways and cause unintended negative effects, despite best efforts to specify the right goals.

• High frequency trading AIs would accidentally cause an economic collapse.

• Humanity would eventually rely on AIs to do everything, which will cause humanity to be more inept for each passing generation.

What is the view of experts on how difficult it will be to ensure advanced AGI systems are beneficial?

• It will be straightforward to ensure AGI is beneficial

• It will require a moderate amount of research to ensure AGI is beneficial

• It will be an extremely difficult challenge to ensure AGI is beneficial

• We do not know