AI optimism has to answer to the people living with the consequences. Before leaving to film America Vs AI, Andrew DeGood and Liz Short debate what they expect to hear on the road and what they owe the people they meet.

Andrew wants the conversation to move from anger toward useful outcomes. Liz pushes back on treating legitimate concerns as pessimism. The disagreement sharpens when the hosts ask whether promised benefits mean anything before people can see them in their own lives.

In this host-only episode:
• Why unscripted conversations matter more than predictions about public opinion.
• How fear, communication, and trust shape the AI debate.
• Why optimism needs evidence and skepticism needs room to speak.

The trip begins with a question, rather than a verdict: why do people feel the way they do about AI?

Follow Ghost in the Machine and bring your answer to the conversation.

Why do AI projects fail before they create value?

Andrew DeGood and Liz Short trace the problem past the model and into the organization. Leaders buy broad promises, ask busy employees to carry implementation, and make technology decisions without the people who understand the work. That can mean chasing automated underwriting while processors still move files between systems and repeat basic clicks all day.

The conversation turns when Andrew asks what a CEO is actually responsible for during an AI transition. The answer is culture. Executives should understand the technology, but their first job is to communicate clearly, involve frontline experts, and give employees a reason to trust the change.

The practical strategy is smaller and more human. Pick one painful workflow. Bring the best person doing that work into the room. Solve the problem, prove the value, and build from there. AI adoption works when leadership starts with people.

When AI leaders agree that the race needs brakes, should people trust the warning or question the incentives behind it?

Andrew DeGood and Liz Short test the extinction case without treating fear as evidence. They examine a former Anthropic researcher's warning, the industry consensus, open-source competition, recursive self-improvement, and OpenAI's use of thousands of agents on the Navier-Stokes problem. The strongest conclusion is not that catastrophe is impossible. It is that extraordinary claims still need precise facts.

Then the conversation turns from hypothetical risk to present power. Andrew argues that personal privacy is already gone. Liz pushes back: surrendering privacy hurts people with less power first, especially when surveillance tools can be abused by institutions or individuals.

The episode lands on a harder standard for both optimism and skepticism. Ask what a system can do, who controls it, what evidence supports the claim, and who pays when the guardrails fail.