OpenAI Altman: Jobs That Can Be Replaced by ChatGPT Aren’t Real Jobs
"Your Job Today Might Not Actually Be Real Work"
This striking statement comes from a recent conversation between Sam Altman and Rowan Cheung.
In this 30‑minute interview, Altman covered a wide range of topics:
- GPT‑6 development
- Whether ChatGPT might become an “American WeChat”
- Changes in AGI concepts
- Future AI interaction modes
- His reaction to becoming a viral “Sora” meme
The discussion balanced light‑hearted moments with serious insight into cutting‑edge AI technology.
Below is the organized transcript with filler phrases removed for readability.
---
Dev Day 2025 Highlights
Q: What was most exciting from Dev Day 2025?
Sam Altman:
- Excited about apps inside ChatGPT — something I’ve wanted for a long time.
- Even more excited about what people are building with Agent Builder and Agent Kit.
- If I had to choose: running apps directly inside ChatGPT.
---
Q: With 800M weekly users, ChatGPT is now a distribution platform. How should developers use the Apps SDK?
Altman:
- We need iterations to learn how people will use features: call apps by name, or have ChatGPT recommend them?
- Developers may discover new organic distribution methods.
- Surprises always happen after release.
---
Q: You also released docs to help apps get recommended?
Altman:
- Yes, with the disclaimer that products evolve quickly — we’re learning together.
---
Agent Builder Evolution
Q: From GPT Builder to now — what’s changed?
Altman:
- Models are much stronger — the gap between then and now is enormous.
- Learned what users want: build in ChatGPT and deploy externally.
- Building sophisticated systems is now drag‑and‑drop simple, with minutes to deploy.
---
Q: Can you build with zero code?
Altman:
- Yes — even non‑coders can build agents.
- Coders can still do more complex builds.
- This is essentially a no‑code revolution for agents.
---
Q: What does this mean for entrepreneurs?
Altman:
- Software development cycles are dramatically faster.
- More ideas can be tested quickly.
- Not sure yet about all the impacts.
---
Billion-Dollar Agent Companies
Q: When will we see the first billion‑dollar company run solely by agents? Is tech ready?
Altman:
- Not yet, but speculation includes “one‑person” and “zero‑person” companies.
- A few years away, but the fact we can talk credibly about it is incredible.
---
Q: Agents that work for a week without feedback — when?
Altman:
- Codex can already complete a day’s tasks.
- Week‑long runs are coming soon.
- Bottlenecks: smarter models, longer contexts, better memory.
---
Strategic Advice for Building AI Products
Key takeaway:
- No universal formula for a unique edge — must be specific to your product & timing.
- “Make tactics into strategy” — start with what works, adapt from there.
- Memory became a major edge for ChatGPT unexpectedly.
---
Q: GPT‑6 advantage strategies?
Altman:
- Fundamentals haven’t changed: network effects, brand, marketing, data.
- Tactics evolve, but core strategies remain.
---
AGI & Scientific Discovery
Definition: Surpassing humans on most economically valuable tasks.
Altman’s focus:
- AI making novel discoveries — early but significant.
- Examples: small scientific breakthroughs, new methods, problem‑solving in math.
- Predicts society will adapt quickly, as with the Turing Test milestone.
---
AI Work Quality: Avoiding "Workslop"
Workslop = low‑value AI output causing more work through cleanup.
Study findings:
- 41% encountered workslop from colleagues in the past month.
- Cleanup costs nearly 2 hours and $186 per employee/month.
Altman:
- Humans produce workslop too.
- Efficiency gains will win in the market.
- Steep learning curve, but adoption can be rapid with the right workflows.
---
Sora, Parody Videos & Watermarks
Altman on viral Sora videos:
- Initially odd, but quickly adapted.
- Flood of watermark‑free, realistic AI videos is inevitable — better to release early with guardrails.
Goal of Sora:
- High‑quality video aids AGI development.
- Video enables new user experiences and supports spatial reasoning in world models.
---
Revenue Sharing in Cameo-Like Features
- Pattern emerging: short, private share videos outnumber public ones.
- May enable paid generations with revenue sharing for likeness use.
- Too early for conclusions — only launched days ago.
---
Ads & Monetization
- Ads possible if consumption is feed-based.
- Subscription or charge-per-gen if usage is private.
- Final model will depend on usage trends.
---
The Nature of Future Work
Altman:
- Past examples (farmers vs. internet jobs) show “real work” perception changes over time.
- Future work may be easier but human drive will remain.
- Hopes for broad exploration — including space.
---
AI Regulation as Global Policy
Altman:
- Global framework needed for safety as models become very powerful.
- Would aim to reduce catastrophic risks.
---
ChatGPT vs. “American WeChat”
Altman:
- Not aiming to be a WeChat clone.
- Goal: a great AI super assistant.
---
Feature Integration Decisions
Altman:
- Sora is separate because ChatGPT is personal — avoid awkward mix with social entertainment.
- Some features still integrated.
---
Most Promising Agent Applications
Altman:
- Replicate Codex‑level experiences in law, finance, and other industries.
- Long‑term vision: start a company by conversing with agents.
---
Voice as Interaction Mode
Altman:
- Voice is natural in many contexts, but not the ultimate form.
- Past smart speakers failed due to weak AI/infrastructure.
- Ideal: tell a computer exactly what you want, it delivers, then steps aside.
---
Voice Product Development
- Will take time and patience.
- Exploring new device concepts with extremely high quality.
- Not ready to share details yet.
---
Reference Links
- https://www.youtube.com/watch?v=zwnVUiwObl8
- https://futurism.com/artificial-intelligence/sam-altman-real-work-ai
- https://x.com/rowancheung
---
Related Platform: AiToEarn
Platforms like AiToEarn官网 enable:
- AI content creation
- Simultaneous cross‑platform publishing to Douyin, Kwai, WeChat, Bilibili, Xiaohongshu, Facebook, Instagram, LinkedIn, Threads, YouTube, Pinterest, X/Twitter
- Analytics & AI模型排名
- Open‑source monetization ecosystem for rapid deployment & scaling
More info:
---
Summary Insight:
As Altman describes, we’re entering a world of fast‑deploying agents, novel AI discoveries, and evolving interaction modes. The challenge now is efficient monetization & distribution — a space where open‑source ecosystems like AiToEarn may help creators turn AI ideas into sustainable, multi‑platform products at scale.
