How to stop AI going rogue #AI potential risks#AI safety measures#AI benefits#listening#speaking#C1







Artificial intelligence is improving so fast that no one knows what it might be capable of. It brings huge opportunities, but also huge risks. Arjun Ramani, The Economist's global business and economics correspondent, explains what could go wrong. In this lesson, you're going to discuss potential risks and benefits that artificial intelligence can bring along with it.




00:00 - How could AI go wrong? 01:12 - What are the risks? 03:11 - How to practise AI safety 04:42 - What are the benefits?


TASK 1 : How could AI go wrong? What are the risks? - listen to the first and second part of the talk about the potential risks and fill in the missing words. 

1.There's a risk of ........................, couse you can ask a model to say something which is untrue, but in a very sofisticated way and post it allover social media. 

2.There's a risk of ..................so that they might spew harmful content about people of certain classes. 

3.You might ask a model to produce something from a factory setting that requires a lot of energy for and in service of that goal of helping you your factory production it might not realise that it's bad to ............ into energy systems that are connected to the Internet and because it's super smart it can .................... our security defences. 

4.Bad actors might just ................... these models, for example terrorist organisations might use large language models to hack into government websites or produce biochemicals by using the models to kind of discover and design new dgrugs. 


TASK 2 : Listen to the third part of the talk about safety and answer the questions. 

1. What is so called " reinforcement learning from human feedback" (rhfs) ? 

2. What is red teaming? 

3.What are the other safety measures suggested by the man speaking? 


TASK 3 : Watch the fourth part of the interview and match the expressions as you heard them : 

 AI comes with truly                are endless

 possibilities here                    could potentially be helped by these types of models 

 any task                                  tremendous promise





Komentarze