Lessons Learned From Nvidia: Fail, A Lot.
For a couple of months now, it seems like Nvidia is what everyone talks about. To Wall Street, Nvidia is the golden cash cow that brought them 10x or even 100x returns. To startups and technology companies it is a role model and the forerunner to catch up to. Briefly rising above Apple’s market cap last November, Nvidia is the driver of growth (and recent decline) in the Magnificent 7 stocks.
Nvidia Research: The Key Driver of Success at Nvidia
The reason for Nvidia’s recent success and fame can mostly be attributed to their high-end chips that are ideal for AI model training and running. The Blackwell Ultra, announced earlier this week at the GTC conference, is an example of these high-end chips. The software that comes along with Nvidia products are also what drive demand for their GPUs.
The birthplace of these innovative products is Nvidia Research: Nvidia’s research department. While relatively small compared to that of comparable companies, Nvidia Research has birthed many crucial technologies such as ray-tracing, a rendering method that simulates the physical behavior of light, and NVLink & NVSwitch, which allows GPUs and CPUs to communicate at hyper speed.
Bill Dally, chief scientist and senior vice president of research at Nvidia, says this success can be attributed to failing a lot. He was quoted saying “if everything you do succeeds, you're not swinging for the fences. You're bunting.”
Know When to Quit and When to Go All In
Dally says that the best researchers are the ones who have great ideas, pursue them, and are wise enough to quit early if it doesn’t work out.
The reason these micro-failures are so important is because they are cost-efficient while allowing for innovation and creative thinking. Trying out many ideas and abandoning the ones that don’t work out make way for a natural selection of ideas. The strong, viable ideas survive, while failed ones go extinct. This kind of continuous trial-and-error is a crucial mindset in startups too. It also leads to more innovation, as it gives far-fetched, seemingly impossible ideas a chance. An example of this in Nvidia is its deep learning super sampling (DLSS) software which leverages AI to enhance game visuals. The first version, released in 2019, did not provide great results and brought along a lot of skepticism. Fast forward 6 years, DLSS 4 now drastically improves game visuals in even the most high-spec games like “Cyberpunk 2077”
Many startups start with one great idea. We all want to develop our ideas into the next Facebook or the next Airbnb, envisioning something that will change the world. Despite our initial hopes, this isn’t always the case. Things might crash and burn, or even worse, not even takeoff. But as a co-founder myself, I know how hard it is to abandon an idea that so much work was put into. So we tend to keep pursuing the same idea, hoping for the best. When this happens, we should be mindful of Nvidia Research’s lesson on failures and why they are so great.
How Lucid AI Pivoted from Failure
My startup, Lucid AI, also went through a similar process. Our first service, which was drastically different from now, was using AI to find social media users who were discontent with a particular product and asking them for a user interview. We were confident that being able to capture raw opinions from social media and provide companies insights through those opinions would be very helpful. However, we soon learned that many social media users didn’t want to bother doing user interviews and that the need for user interviews were in decline due to advanced insights tools.
After realizing this, we quickly decided to pivot to a new idea while maintaining our AI model and unique advantages that come from it. This is what lead to Lucid Agents.
I believe that if we hadn’t realized that our initial idea was gonna be a failure, or kept on pursuing it despite realizing, we would not be where we are right now. The courage to abandon our idea no matter how much effort put in allowed us to save time and our scarce resources.
How You Can Fail Successfully
No matter how much you think about this in your head, it’s evidently hard to actually let your idea die and pursue another one. I wanna share three measures you can take to make sure you can look at things from a more objective point of view and fail successfully.
1. Set Clear KPIs and Monitor them Regularly
KPIs, or Key Performance Indicators, are crucial when measuring the success of your ideas. Set clear objectives and metrics you need/expect to achieve, checking regularly if your idea is reaching them. This allows you to monitor your performance in a more objective way, helping you make better decisions
2. Don’t Get Hungover on Past Efforts
Don’t fall for the fallacy of sunken costs. Be more future oriented. No matter how much effort you’ve put in, a failed idea is a failed idea. Learn to quickly move on and pivot to a new, hopefully better, idea.
3. Develop Processes that Allow for Quick Testing
In order to know if an idea will fail or succeed, it’s crucial to have a framework that allows for quick testing of ideas. Rapidly test your ideas through prototyping or market research, saving the trouble of getting your idea to market only to realize there’s no demand for it.
