Individual Orientation: Epistemics, Careers, Engagement | The AI Transition Tutorial

The Question

You've read ten chapters on a fast-moving, contested, high-stakes field. What are you supposed to do with it?

The honest answer depends on your situation: your career, your skills, your resources, your values. But some general patterns apply. This chapter is about those.

Epistemic Practices

Before "what to do", the question of "how to think". The AI transition has specific epistemic challenges.

Hold views loosely

The field updates quickly. Views that were reasonable in 2023 have aged in ways that were hard to predict. Views reasonable today will probably age too. Hold specifics loosely; frameworks last longer.

Update on evidence, not tribe

When new information arrives, update your probabilities in response to the information, not your camp's preferred conclusion. This sounds obvious; it's easy to fail. Notice when your updates move with your tribe rather than with evidence.

Separate "is" from "should"

Descriptive questions (is AI progressing fast?) are different from prescriptive ones (is AI progress good?). Confusing them produces bad reasoning. Practice holding them separately.

Notice overclaiming

Certainty above what evidence supports is a marker in any direction. "AGI by 2027", "AGI never via LLMs", "alignment is trivially solvable", "alignment is impossible": all require humility the statements don't show.

Check arithmetic

AI arguments often involve numbers (compute, cost, capability, probability). Doing a minute of arithmetic on a claim disproportionately improves your reasoning. Many arguments collapse when numbers are checked.

Watch your emotional state

Strong emotion (excitement, fear, disgust) degrades reasoning. If you're having strong feelings about AI, notice, and try to form views in calmer moments.

Career Decisions

If you're choosing or evaluating a career with AI context, some considerations:

The dominant uncertainty

Your career exists in a range of possible futures:

Slow transition: AI is a useful tool, work continues much as before
Medium transition: significant automation in some fields, gradual adjustment
Fast transition: rapid transformation of many careers
Very fast transition: careers as we know them end within your active working years

Your career strategy should hold up across several of these. If you're betting everything on one specific trajectory, you're betting on something you can't predict.

What probably transfers across futures

Skills and dispositions that hold value across most plausible AI trajectories:

Judgment: the ability to decide well under uncertainty and ambiguity
Interpersonal skills: most work still involves humans; this is unlikely to stop mattering
Deep expertise in something: broad shallow skills are easier to automate; deep specialised ones are harder
Ability to use AI well: if AI is a significant tool, using it well multiplies your output
Adaptability: the raw capacity to learn new things
Integrity: trust is valuable in every environment; more so in environments where AI is generating plausible-seeming falsehoods
Systems thinking: understanding how things fit together at organisational and societal scale

What probably doesn't transfer well:

Narrow technical skills without depth
Tasks AI is already doing competently (simple writing, basic coding, data entry)
Credentialed work whose value comes from restricting access

AI-adjacent careers

If you work on AI directly, the field has roles across:

Capability research (at labs and academia)
Alignment and safety research
Policy and governance
Ethics and responsible AI
Applied AI (specific industries)
Infrastructure (chips, data centres, data)

Each has different impact profiles and ethical questions. Before optimising for your career in the field, think about what you want the field to become.

Non-AI careers in an AI world

Most people won't work on AI directly. They'll work in a world shaped by AI. For this, think about:

What do I produce that AI does worse?
How does AI change the shape of my work?
Where will value concentrate in my field?
What skills should I develop over the next decade?

A medical professional, teacher, lawyer, writer, or engineer can answer these specifically for their field; generic answers are less useful.

Don't over-optimise

A trap: optimising your career entirely around an AI prediction that may be wrong. Bet modestly on likely outcomes; don't commit irrevocably to one scenario.

Financial Decisions

Related but separate from careers.

Investing in AI

If you're investing:

Infrastructure (chips, data centres, energy) has benefited; continues to
Applications vary; many AI products have weaker economics than advertised
Frontier labs are mostly private; public exposure is indirect (Microsoft, Google, Meta, Nvidia, Oracle)
Timing is genuinely hard; bubble dynamics exist; exit points unclear

Standard disclaimer: AI-related investing doesn't escape normal investing rules (diversification, risk management, long horizons). Enthusiasm for AI doesn't exempt you from discipline.

Saving for an uncertain future

If transformative AI arrives soon, your savings may need to last differently than traditional retirement planning assumes. If it doesn't arrive, standard planning still applies.

A prudent response:

Save meaningfully; don't assume future-you has lots of help
Keep some flexibility; don't lock into illiquid bets on specific futures
Maintain useful skills; don't assume the work you do now lasts forever
Watch for opportunities that specifically arise from AI, but don't require them

Insuring against tail risks

Some low-probability-high-consequence outcomes of the transition warrant thought. Physical disasters from infrastructure failure. Economic shocks. Social unrest. These are small probability, large consequence.

Some standard prudence covers much of this: emergency savings, diversified networks, meaningful relationships, health. Nothing specifically AI about it; AI just adds salience to the general case for preparedness.

Engagement Decisions

Whether and how to engage publicly on AI.

The engagement options

Privately informed: you know what's happening; you don't write or speak publicly about it
Active learner: you engage with the discourse as a reader and thinker
Participant: you contribute your own writing, analysis, or research
Activist: you push for specific outcomes through organising, lobbying, or campaigns
Professional: your job is AI-related in some way

These are increasing commitments. Each is legitimate.

When to engage more

Good reasons to engage more publicly:

You have specific expertise relevant to the debates
You're seeing framings you think are misleading and you can articulate why
You have specific policy ideas you want advanced
You have a platform that can carry useful signal

Less good reasons:

You feel strongly and want to vent
You want to be part of a hot topic
You're angry at a specific person or lab

The downsides of engagement

Public AI discourse is often unpleasant. Real trade-offs:

Time lost from other valuable work
Exposure to harassment
Pressure to stake out positions you'll later want to revise
Being associated with camps you don't fully endorse
Epistemic distortion from the incentives of engagement

A reasonable person can reasonably choose to engage little. "Not public on the internet" is a valid position.

If you do engage

Engage with ideas, not people (mostly)
Update publicly when you change your mind
Cite sources; specify where you're uncertain
Don't share slop; check before amplifying
Know what you don't know

Using AI Responsibly

Whether or not you engage publicly, you probably use AI. A few principles for users:

Don't trust blindly

Verify important outputs. AI hallucinates; checking isn't paranoid, it's literate.

Don't outsource judgment

Use AI for drafts, suggestions, analysis. Make the decisions yourself, with your values and context.

Maintain skills

Don't let AI do things you want to keep being able to do yourself. Writers who don't write without AI lose writing ability. Programmers who don't code without AI lose coding ability. Choose what to retain.

Consider externalities

AI use has environmental costs (compute) and sometimes labour costs (data labellers, moderation workers). Using AI isn't ethically neutral; it's not prohibited either. Hold the costs in mind.

Notice dependence

If you feel genuinely anxious without AI assistance, notice. Some of this is fine (I feel anxious without electricity too). Some may suggest over-reliance.

What to Actually Do

For a person who wants a small list of actionable things:

For epistemics

Pick one source from each camp you don't naturally read, follow for a year
Write down what you currently believe; revisit quarterly
Do arithmetic on numeric claims before trusting them

For career

Identify one skill that probably transfers across AI futures and invest in it
Identify one way AI might reshape your specific field and think through your response
Don't bet your career on a single timeline forecast

For engagement

Pick one engagement level and stick with it for a year
If active, keep personal notes on your evolving views
Notice when engagement is costing more than it's producing; adjust

For using AI

Establish some domains where you don't use AI (to preserve skill)
Establish some where you use it aggressively (to gain productivity)
Be honest with yourself about which is which

The Background Reality

One closing note. All of the above assumes the future is uncertain and our actions matter modestly. Both are true. The discourse often encourages either (a) despair ("it's too late to matter") or (b) messianic urgency ("you must act now in this specific way"). Neither is warranted.

A sensible position:

The future has real uncertainty; your plans should hold up across scenarios
Your individual actions contribute to outcomes; not decisively, but meaningfully
Cultivating the habits of careful thinking, honest engagement, and preserved human agency is valuable regardless of how AI develops
Much of what matters in life is not about AI; keep perspective

Common Pitfalls

"I'm just one person; nothing I do matters." Many small contributions aggregate. Pessimism about collective agency is itself a choice

"I need to go all in on AI." Probably not. Over-commitment to one scenario is expensive if that scenario doesn't arrive

"I'll just opt out." You can try. You won't fully. AI is increasingly integrated into systems you use. Literacy is valuable even for opt-outs

"I should follow the experts." There isn't an expert consensus to follow. Forming your own informed view matters

"If I'm not worried enough / worried too much, I'm doing it wrong." The appropriate emotional state is hard to specify. What matters is action that suits the actual situation, not matching your emotions to a reference level

Next Steps

Continue to 12-best-practices.md for habits that hold up through years of change.