1 of 11

"We propose that a 2-month study

by 10 men can solve AI."

— Dartmouth Conference Proposal, 1956

Were they right?

CSE 5 – Discussion Session

2 of 11

Event 1 of 5 — Which Big Idea?

CSE 5 – Discussion Session

1997

Deep Blue beats the world chess champion

Big Idea: Representation & Reasoning

IBM's Deep Blue evaluated 200 million positions per second using hand-crafted rules — and won. But it could only play chess. Ask it to play checkers and it was useless.

💬 Why can Deep Blue beat a world champion at chess, but still not "think"?

3 of 11

Event 2 of 5 — Which Big Idea?

CSE 5 – Discussion Session

2012

AlexNet cuts ImageNet error rate in half

Big Idea: Learning + Perception

Error rate dropped from 26% → 15% in a single year. The secret wasn't a smarter algorithm — it was GPUs and 1.2 million labeled images. Data + compute = breakthrough.

💬 What changed in 2012? Was it the algorithm, or something else?

4 of 11

Event 3 of 5 — Which Big Idea?

CSE 5 – Discussion Session

2016

AlphaGo defeats Lee Sedol

Big Idea: Learning

Unlike Deep Blue, AlphaGo was given no rules. It learned by playing millions of games against itself. Lee Sedol, one of the world's best players, lost 4–1.

💬 If AlphaGo learned from scratch with no human rules, what does "learning" really mean here?

5 of 11

Event 4 of 5 — Which Big Idea?

CSE 5 – Discussion Session

2018

Amazon's AI misidentifies 28 members of Congress as criminals

Big Idea: Societal Impact

Amazon's face recognition tool incorrectly matched Congress members to criminal mugshots. The training data carried human bias into the system.

💬 This system was trained on real data. So who's responsible when it gets it wrong?

6 of 11

Event 5 of 5 — Which Big Idea?

CSE 5 – Discussion Session

2011

Siri launches — people love it, then get frustrated

Big Idea: Natural Interaction

Siri could answer questions, but it missed context, misheard accents, and broke on anything ambiguous. Natural language is hard — it's packed with cultural assumptions and unstated meaning.

💬 Why is talking to a computer naturally so much harder than searching Google?

7 of 11

8 of 11

AI Tools — Where Are We Now?

CSE 5 – Discussion Session

9 of 11

Survey – What’s your interests?

CSE 5 – Discussion Session

10 of 11

AI tool introduction – Tool Card

CSE 5 – Discussion Session

  • Explain the relevant terms

  • (a little bit) technical details to how it’s made (trained)

  • Demo

  • How to use it

  • How it could be related to your field (e.g., AI for xxx)

  • Try on your own

11 of 11

Preview – How to write a good prompt?

CSE 5 – Discussion Session

1

Give it a role + context

Who is it? Who are you? What's the situation?

2

Specify the output format

Length, structure, tone, what to include/skip.

3

Give an example (few-shot)

Show it what good looks like — don't just describe it.

4

Iterate, don't rewrite

Good prompts are conversations, not one-shots.

5

Tell it what NOT to do

Constraints are as powerful as instructions.