LLMs Need More User Experience Wins to Avoid the Silicon Valley Graveyard
Nikola Tesla is widely recognized as the father of modern electricity. Yet, he spent later part of his life here in New York City with very limited resources. Much of the profits generated from his inventions was made by businessmen who created consumer experiences, such as electric light bulbs, or developed distribution systems that brought his work to the masses.
As we talk about AI as the “new electricity,” this historical lesson from the early days of modern electricity is crucial to remember.
If you are building a consumer product that leverages large language models (LLMs) to deliver new experiences without fine-tuning, you’ve likely heard this criticism: “It’s just a GPT wrapper.” As someone who has built both AI models and consumer products, the disdain for consumer applications using this transformative technology is surprising. While it is true that LLMs represent a groundbreaking advancement, building consumer experiences that resonate with and delight users is far from trivial: roughly 80% to 90% of such efforts ultimately fail.
Revolutionary technologies we recognize today only saw mass adoption through simple, user-friendly consumer experiences:
Internet and the Web Browser: The internet was invented in the 1970s, but it became a household phenomenon only through the introduction of web browsers like Netscape and Internet Explorer, which made it accessible to the average user.
Smartphones and Mobile Computing: The core technologies behind smartphones — wireless internet, touchscreens, and advanced processors — existed for years. However, it was the release of the iPhone, with its intuitive interface and app ecosystem, that made smartphones indispensable.
GPS and Consumer Navigation Devices: Originally developed for military use, GPS technology gained mass appeal through easy-to-use handheld devices and applications like Google Maps, which simplified navigation for everyday users.
Personal Computers and Graphical User Interface (GUI): The computing revolution took off with user-friendly personal computers like the Apple Macintosh and later Microsoft Windows, which made computers accessible to non-experts through intuitive GUIs.
Wireless Networking (Wi-Fi) and Easy Setup Routers: Wi-Fi became popular through the development of consumer-friendly routers that simplified internet access without complicated wiring.
Digital Photography and Point-and-Shoot Cameras: Digital image sensors became mainstream when they were incorporated into simple point-and-shoot cameras, making photography accessible to everyone.
Streaming Services and Smart TVs: Streaming content became widespread when services like Netflix, coupled with smart TVs, made media consumption simpler and more convenient than traditional cable.
Voice Assistants and Smart Speakers: Voice recognition technology became widely adopted when products like Amazon Echo and Google Home offered an intuitive, hands-free way to interact with devices and access information.
As the post-ChatGPT euphoria fades and companies are pressured to deliver on AI’s promises, it’s an opportune moment to focus on consumer applications that provide tangible value. At PhotoSpot, our AI travel planner, we experimented with both reinforcement learning to enhance travel recommendations from LLMs and improvements in consumer experience. Interestingly, most of our success has come from refining the user experience rather than upgrading the model’s recommendations.
For those who have been in Silicon Valley long enough, the Gartner Hype Cycle feels accurate when reflecting on the waves of hype we’ve witnessed over the years. If LLMs want to avoid joining the ranks of failed technologies in Silicon Valley, it’s essential to back companies that can create true consumer value
