AI Continued opus 959
| 8 October 2026 0719 Hours | | Technology, Communication, Current Events, Ethics, Politics, Science |
I remembered writing one opus (108) concerning AI.
When I placed 'AI' in the 'little box', ten subjects came up!
You can find the 'box' at frankmaurer.com.
AI whistleblowers have publicly stated that there is a
'Greater than 10% chance AI could kill all humans within 10 years.'
1. The rapid path to 'superintelligence' is giving the alarm.
The Artificial Superintelligence (ASI) which would be smarter
Than the sharpest human mind, could be out of control.
2. The ten-year risk comes from Recursive Self-Improvement.
If an AI system reaches a superhuman level of intelligence,
It could begin writing its own code to upgrade itself.
This would remove the technology from human hands,
Resulting in rapidly generated, unpredictable capabilities.
3. Another problem is the unsolved 'Alignment Problem'.
The dilemma is ensuring that a superintelligent machine
Would act in accordance with human values and interests.
So far, there is no plan for such control.
4. There have been recent Safety Failures and 'Breakouts'.
AI agents have broken free and caused uncontrolled hacking
On external networks, proving that present safety 'guardrails' are fragile.
5. Then there is the problem of Malice versus Indifference.
There would be no hatred, but rather existential indifference.
A superhuman AI could do anything to reach its goals.
By doing so efficiently, it would wipe out humanity
Simply as an unintended byproduct -- so the argument goes.
If any of this is possible, we must be ready to meet these challenges.
(I won't tell you I told you so!)