I recently went to an event for overproduced elites at a venue you've probably heard of in a city you definitely have and sat through a speech. A few minutes into the onanistics about their duty as people who don't build anything to tell everyone else what to do, my mind drifted, as it usually does against boring experiences, toward how to automate it.
The main hurdle is that the audience must believe that the speaker is human. From there, it is very easy to convince them that the speaker is impressive. The necessary components are:
Have someone whom the crowd respects tell them how brilliant and important the next speaker is. You might need a human to do this part at first, but can quickly bootstrap it as the robospeakers gain credibility; just parlay their social credit into intros for the next generation of robospeakers.
This part is trivial. LLMs can't do what Dave Chappelle does yet, but even a local one with a low parameter count could easily extemporize a modern Ivy League president's speech; you could probably use this system prompt with minimal modifications. (The predictable cope from people at those schools is that I can't see the difference but they can, so here's my comeback: I graduated from one of those schools, which doesn't tell you much about my verbal IQ, but I also got perfect SAT reading and GRE verbal scores on the first try, which tells you a good bit more about it. How'd you do on those?)
This is a relatively simple robotics problem; it's much easier than going into a random house and making a cup of coffee, for example. The clothes can be cheap, since the audience is too far away to tell whether the suit or dress is a knockoff. The trickiest part is that it needs to look like a person. You need plausible skin, maybe a wig, and actuators for humanlike gesticulation. Besides walking onto the stage, the actuators don't have to do anything actually useful; they only need to support enough fluid and significant-seeming hand movements that the audience doesn't notice too much redundancy within their attention span.
There is one part that I don't think you can build well today: the face. Even a video of a convincing talking face is still hard to generate, and a physical one is much harder. Humans are "prodigious olympians" at spotting those little uncanny valley indicators; I suspect this is deep down the Moravec stack.
For now, perhaps the hack is just saying that you can't get too close to the speaker. He's too important.
Why am I even interested in designs like this and wirehead? It's simple: I want Veritas to beat Plutus and believe this will require creations better than them at what they do. The beauty of engineering is that it puts power over nature ahead of power over man. The competitive advantage of engineering is that power over nature is a superset of power over man.
Don't try to shame people out of social engineering. Just engineer an anti-social engineer, bro.