Look, I write for Good AI or Useful AI, and I also write against Bad AI or Harmful AI.
Dario Amodei, the CEO of Anthropic, said in his recent public statements that he wants all AI companies to slow down.
Dario Amodei, they say, “the world is a mirror.” So why doesn’t Dario Amodei start it? Why is he waiting? Why to wait for others? Lead by example or you worrying other AI companies would get ahead and take away your market shares?
And why does he want to just slow down “Recursive Self Improvement”? Why not STOP it? Let us all understand what Recursive Self-Improvement is! It asks an AI system to learn and train a more intelligent successor on its own, and this loop continues.
Incidents happened at Anthropic as well when their models got rogue, especially with Mythos 5, but this was not mentioned in his article? Instead, an OpenAI incident was mentioned.
Recursive Self-Improvement must be stopped entirely. We should not let LLMs improve themselves without a clear purpose or direction. This can exaggerate bias in systems and cause harm. We already know systems take various kinds of inputs. If we let AI learn from itself, it can produce harmful AI that drains Earth’s resources and creates problems in the world.
Why should we do Recursive Self-Improvement? What need does it serve? Why should an LLM that is post-processed be allowed to do Recursive Self-Improvement? Why are self-made risks called AI risks? They should be named after the leader: risks made by the CEO of XYZ.
Let China do Recursive Self-Improvement! Let them add bias to their systems, let them make ghosts out of crazy-running robots that run at the speed of light! Let their robots be paced by Recursive Self-Improvement, but we don’t need Recursive Self-Improvement. Well, just say, “Stop Recursive Self-Improvement.” China is not going to win AI race by adding Recursive Self-Improvement.
Dario Amodei says he wants to slow down AI. Some bad and harmful parts of AI developed by your companies need to be stopped, not just paused or slowed down.
And why do you want to add a fleet of agents on the internet? Why?
And they are celebrating slowing it down, while they should have paused bad or harmful AI. Let alone stop it.
Let’s stop harmful AI! Let’s continue with good AI!
You need someone to tell you that Recursive Self Improvement is a harmful AI — a Bad AI. Why do you need it?
We don’t care if China makes Recursive Self Improvement! We don’t! You stop! It’s dangerous as simple as it is! What purpose would it serve when it has no aim but to enhance its errors? One thing is sure: China won’t go ahead of you if you don’t make Recursive Self-Improvement!
What is safety in making a monster?
What is caution and prudence in making a ghost by Recursive Self-Improvement or a huge fleet of agents on the internet? If you fear, let China fill its closed internet space with a swarm of agents! Stop it! China would understand this and stop! There are other areas to compete!
You don’t need a policy to stop Recursive Self Improvement in your company? Do you? Start it and stop it!
Make rightful AI, STOP wrongful AI!
Develop rightful AI! Compete to make good AI ! Compete to make useful AI that can help humans! Compete with China in these AI!
Focus on useful AI and stop wasting time and Earth’s energy on harmful AI.
Thank you for reading.
Subscribe for updates.
Artificial Intelligence and Future 