Artificial Intelligence – Doomsday Debate

The robots are not coming to kill us.

• History is replete with concerns that new technology would cause great harm or bring about the end of the world, but we are still here, and I suspect that AI is no different.
• Thanks to a “whistleblower” or two who presented no evidence of what it is that they have seen, statements from the frontier labs that are deeply conflicted and general agreement from those that have most to gain, the doomer debate is raging once again.
• It is important to note that this has been going on for over 2,300 years, and I suspect that, as usual, history will end up being right.
• Here are a few (and in some cases, amusing) examples:
  • Writing, 370BC: where Plato composed a dialogue where Socrates describes the Egyptian King Thamus’ rejection of writing.
  • King Thamus feared it would breed forgetfulness and leave people unable to defend their own ideas.
  • This is the oldest known panic over a new technology.
  • Printing Press, c1490: where the Benedictine abbot Trithemus argued that the printing press threatened the religious and intellectual culture built around scribes.
  • He was right about the scribes but completely wrong about its impact on religious and intellectual culture, which is far more important.
  • Coffee, c1670: where the 1674 “Women’s Petition Against Coffee” claimed it made men impotent.
  • If this was true, then the human race may have become extinct, and Silicon Valley would never have existed.
  • Vaccination, 1802: where a cartoon went the 17^th-century equivalent of viral, showing humans sprouting cows after being vaccinated with Jenner’s cowpox vaccine.
  • The clergy then went on to call it unchristian to inject animal matter into humans.
  • Railways, c1820: where it was claimed that speeds of over 20mph would suffocate passengers (probably apocryphal) or cause brain damage.
  • It was also claimed that “railway spine” and “railway madness” (not clear what either of these are) would afflict passengers.
  • Bicycle 1890s: where doctors warned of “bicycle face” or that cycling would lead to general weakness of mind, lunacy or homicidal mania.
  • Some might argue that, looking at today’s urban cyclists, there is some element of truth to this, but it has taken over 100 years to manifest itself, meaning that it is probably not the cycling that caused it.
  • Atomic Technology 1950s: where there was concern that an atomic explosion would ignite the atmosphere and kill everyone.
  • There is no doubt that atomic technology has the potential to wipe out mankind, but this is down to misuse by humans of the technology, not the technology itself.
  • Particle Accelerators 1999-2008: where it was thought that particle collisions would open mini-black holes that would consume the Earth.
  • This was carefully debunked, allowing the accelerators to operate without triggering any of the cataclysms described.
• The message here is that new technology has always created evidence-free panic and fear about the end of the world or cataclysmic consequences, but they have always been either completely wrong or massively overstated.
• I think that this time around it is no different.
• This is especially the case as we have been warned of this multiple times already (GPT-2, o1, “thinking models”, Mythos), all of which have come to nothing.
• In order for the machines to come and kill us all, they would have to gain a deep understanding of causality, which so far, they have demonstrably failed to do.
• They would also have to decide that the world would be better off without humans and have access to the tools required to bump us off.
• Robots can barely make coffee or pick up complex objects, and the same software in an exact copy of the same robot still produces different results.
• Furthermore, the machines are still unable to draw accurate pictures of left-handed tennis players and still seem to be unable to reason from first principles.
• It is mastery of the simple things and not taught skills like PhD maths that will herald the age of superintelligent machines, and of this there is still no sign.
• Hence, I see a lot of this noise as a disingenuous attempt by the big players at regulatory capture to prevent the upstarts and open source from challenging their technological supremacy and degrading their economic power.
• I judge these companies by what they do, not what they say, and so far, their actions are clearly pointing towards revenue and profit maximisation, for which they are strongly incentivised.
• There is nothing wrong with this, but the dishonesty lies in dressing it up as fears of Armageddon.
• The same is true of the whistleblower who 10 days ago was a nobody, but who will now be able to command large fees for speaking engagements.
• The net result is that I see LLM-powered AI as a large technological advance that will greatly increase human productivity.
• Whether there is demand for all of this productivity remains to be seen, but at the moment, there is no doubt that the demand is there for 2026 and most likely for 2027.
• Consequently, I remain invested in the AI trend as the next 15 months at least look very good, and I sleep well at night, safe in the knowledge that the robots are far too stupid to do me any harm.