033 - Outside View

Another useful method of Perspective-Taking was termed "The Outside View" by Nobel Prize winner Daniel Kahneman.

This method integrates critical and often overlooked sources of information, such as base rates related to the subject matter. The alternative that executives often fall into is "The Inside View", where they examine the specifics of a case, but fail to consider "analogous cases and other external sources of information".

Similar cases are critically important to consider in most significant decision-making processes. One my team has encountered several times now is when potential investors make a statement that roughly translates to mean:

  • "I'll be interested in investing when you can demonstrate a system that outperforms the best several multi-billion dollar companies were only able to produce several years after they were funded."

The phrase "putting the cart before the horse" comes to mind. Of course, that also means that they're saying:

  • "I'll be interested in investing when you no longer need my money, there is a line out the door, and it'll cost 10x as much or more.".

Most people can probably agree that both statements sound completely ridiculous, and yet that is the sort of thing I've heard potential investors say several times now. This is what people risk sounding like when they fail to take "The Outside View".

When most AI startups today raised their first million or more, they had either precisely, or very nearly nothing but empty words. This was true of OpenAI and Anthropic, as well as most of the newer contenders. Even raising massive funds and demonstrating virtually no innovation, just copying Google, both of those companies still took 2 years or more to accomplish anything. So, let's compare those analogous cases:

  1. Expectation-1: "Exceed the top performers in an industry, pre-funding, and on my terms."

  2. Expectation-2: "Have either precisely or nearly nothing but empty words, just copy another company."

Again, neither of these expectations is reasonable. The irony between these two is of course that many of the "top performers" in the AI industry wouldn't exist today if they hadn't been greeted with Expectation-2.

Perhaps it is time for Expectation-3: "Demonstrate that you have a firm grasp of what you're doing, some innovation, and a proof of concept."

If such a reasonable bar was set, we could avoid the startups that only emerged from Expectation-2, without preventing virtually all startups from emerging as Expectation-1 could cause.

The choice is left to investors, but the consequences of that choice are something all of humanity pays the price for. Will your investments be a meme?

*Note: The ethical side of this problem is also proportionately more severe, as it is subject to a massive force multiplier in this particular case. That is the subject of a Thought Experiment where a number of often-overlooked factors in ethics are placed in the context of developing more advanced AI systems, like the real-time and scalable version of technology we've demonstrated since 2019: The Ethical Basilisk Thought Experiment.

As noted in the thought experiment, Actions and Responsibility matter and are both subject to this force multiplier, both positive and negative. However, neither beliefs nor intentions matter at all. It is unfortunate for those who remain oblivious and blindly walk off a cliff, but that is reality.