The “@grok is this true” phenomenon
On X users often reply to a post with “@grok is this true” which triggers Elon’s LLM to respond to their post. As UX go it works pretty well. The problem is that people use it for the most mundane questions. It really emphasizes how little the average person knows about things. Even worse people attempt to use “@grok is this true” in online arguments. I’ve also had people attempt to use their ChatGPT Deep research output in arguments against me. This is a problem. Because LLMs will say whatever you want them to. You can’t use an LLM in an argument. An LLM is not a reliable source. But more importantly using LLMs in arguments is bad because it makes arguing with you comparable to a conversation with an LLM. Why bother talking to you if it’s the same as talking to an LLM?
The “LLMs will never achieve intelligence/AGI/my standard” trope
You don’t know that. You cannot prove a negative. No one understands this 5 year old technology well enough to say how far the thousands of engineers and trillions of dollars being spent on it will go. It might get stuck somewhere. It might keep going beyond AGI into the much imagined singularity. You like the rest of us have no idea where this is going. Wait until we go a full calendar year without an improvement coming out in some aspect of LLM based AI functionality.
The “Drone swarms” trope
Drone swarms are not a real thing. 99% of drones used in anger have a human controlling them every second of flight. The remaining drones have a human specify their course and target manually ahead of time. AI is nowhere near as good at flying drones as humans are. FPV drones are terrifying because they allow a human to precisely target an explosive charge remotely. AI is nowhere near that good at flying anything.
The other problem with drone swarms is that they do not make a ton of sense. The advantage of drones is that they allow you to precisely target things with cheap explosives. They are extremely cheap precision guided munitions. The common FPV drone has a maximum flight time of around 30 minutes. That’s not a lot of time for swarming. So if your drone doesn’t find something to hit in that time, it’s falling out of the sky. If you have a drone swarm with thousands of drones, you need thousands of targets for those drones to attack. The battlefield at least in Ukraine doesn’t provide any place with thousands of targets within drone swarm range.
Drone swarms make sense for two types of attacks. They make sense on the first day of a war if you have the element of surprise and are targeting a strategic asset. Basically Pearl Harbor but with drones. Secondly, drone ‘swarms’ make sense for terrorism against civilian targets. Terrorists could generate a lot of terror and have a good chance of living long enough to run away. Won’t be fun for them when they get caught though.