In June, I attended Y Combinator’s Startup School, a series of free online classes that teach you how to build and grow startups at the famed accelerator that helped launch services like Reddit, Airbnb, and Instacart.
But since OpenAI released GPT-4o, the latest version of ChatGPT, in May, the AI has been praised as a tool for every little thing from job interview preparation to homework help. Previous versions of ChatGPT were already able to help with digital marketing and brainstorming business ideas.
Can ChatGPT help aspiring founders overcome their initial hurdles, with a record 5.5 million entrepreneurs filing latest business applications last 12 months and thousands and thousands more on the way this 12 months? And if so, how does it compare to human-led tutoring?
While ChatGPT has not been without controversy (it was accused of getting a “disturbing resemblance” to Scarlett Johansson, and its release raised questions about the protections people have when protecting their likenesses), as GPT-4o becomes available to a wider audience, it could prove useful for aspiring business owners.
In fact, OpenAI CEO Sam Altman was heavily involved in the Y Combinator project, working as its President from 2014 to 2019 and announcing Startup School while he was there.
GPT-4o has a paid version that costs up to $25 per thirty days and a free version, but free users have limits on the variety of messages they’ll send and access to advanced features. I used the free version.
Here’s a comparison of the two methods in answer to some steadily asked questions.
Question: Should I start a startup with no experience?
ChatGPT (free version of GPT-4o): ChatGPT’s answer was more detailed and comprehensible than I expected. Instead of a paragraph, ChatGPT gave me a step-by-step answer with bullet points – a general overview.
Full Answer. Source: OpenAI/ChatGPT
The AI chatbot gave me the pros (learning experience, passion and drive, fresh perspective) and cons (risks, steep learning curve, resource management). It offered steps to mitigate the risks of inexperience, reminiscent of taking courses, networking, and doing research. It also really helpful resources, including the book The Lean Startup by Eric Ries.
Combinator Y: During the first session, titled “Should You Start a Startup?”, Y Combinator’s Harj Taggar said that in his experience, startup founders’ success isn’t determined by what school they went to or how confident they appear — what matters most is their resilience.
Taggar’s advice addresses the emotional heart of the query: “Can I do this without having done it before?” may really mean, “Am I good enough to try this?”
His answer may speak to a deeper sense of imposter syndrome and help someone realize that even if he hasn’t founded 300 firms or built a billion-dollar business yet, he still has what it takes to launch a startup.
How to find a winning startup idea?
ChatGPT: Once again, ChatGPT provided a more practical overview that may very well be a good start line. It included a multi-step response to the query, starting with identifying problems and pain points and ending with assessing the market size and feasibility of the idea. The eight steps were short and open-ended.
Partial answer. Source: OpenAI/ChatGPT
For example, step 6 was titled “Validate your idea” and included the following points:
- “Minimum Viable Product (MVP): Create a simple version of your product to test it in the market”
- “Feedback: Collect feedback from potential users and make changes based on their comments”
- “Pilot Programs: Conduct pilot programs to gauge interest and feasibility”
Combinator Y: Human advice was more specific and helpful. In his class “How to Sourcing and Evaluating Startup Ideas,” Y Combinator partner Jared Friedman gave a clearer sense of the “why” behind winning startup ideas: 70% peak 100 Y Combinator firms, including DoorDash and Stripe, got their ideas by taking notice.
“The problem is that when people sit down and try to come up with startup ideas, they tend to come up with bad ideas,” Friedman said.
Friedman advised would-be founders to work at startups, build interesting things on their very own, and try to gain expertise in a given topic to find the right ideas.
Application
ChatGPT, while it could sound and look more human than ever, is still more of a one-stop web search engine than a business mentor—for now. While it may well be helpful and complementary to human business advice, I’ve found that it’s not yet a alternative for the invaluable experience of learning from an expert, especially a Y Combinator partner.
It has the content you would find in an article about small business ideas or similar, but it lacks the nuance of human answers to the same questions.
According to Dario Amodei, CEO of $18 billion AI startup Anthropic, it currently costs about $100 million to train an AI. Over the next three years, Amodei predicts it can take between $10 billion and $100 billion before models turn into higher than “most humans at most things.”
Meanwhile, AI carries a hidden cost of upper greenhouse gas emissions, although it could play a role in reducing emissions in the future. The vast amounts of coaching data it requires and its potential to reduce human work are also concerns.
For a technology that uses a lot of resources, ChatGPT requires strong use cases. While ChatGPT can write, run ad campaigns, power additional classes, and act as a good start line for answering startup questions, this short experiment shows that it may well’t yet replace human answers and knowledge.
Still, business leaders are optimistic about the technology.
“I’m more excited about AI than I am worried about it,” Sir Richard Branson told Entrepreneur last month. “I think it’s going to change healthcare, it’s going to change surgery, it’s going to change many, many aspects of life.”