Amazon CEO says it has been working on ChatGPT-like tech for a long time

While Microsoft and OpenAI have been soaking up all the AI spotlight, Amazon's top boss finally decided to break the silence.

In an interview with the Financial Times, Amazon CEO Andy Jassy admitted that the possibilities around generative AI are genuinely exciting. But he quickly pushed back on the narrative that Amazon was caught sleeping, claiming deeply technical giants like his have been quietly building these massive generative AI models for a long time.

The reassurance comes as investors grow anxious that Amazon is lagging behind.

The pressure from investors

Amazon already has AI touchpoints in millions of homes and offices, from its voice assistant Alexa launched in 2014 to its coding assistant CodeWhisperer.

Yet Wall Street wants more. Matt McIlwain, managing director of early Amazon backer Madrona Venture Group, told the FT that Microsoft has clearly taken the lead and captured massive mindshare. In his view, Amazon urgently needs a clear strategic response.

Jassy noted that Amazon is actively exploring partnerships with smaller startups. The prime example is Stability AI, the creator of Stable Diffusion and a rival to OpenAI, which recently named AWS its preferred cloud partner to train and build models.

Alexa wants personality, but staff are using ChatGPT

On the consumer front, Amazon knows Alexa needs an overhaul. David Limp, Amazon's senior VP for devices and services, told the Times of London that Alexa is currently far too transactional and mostly just answers isolated questions. The team is now working to make it truly conversational.

Limp framed ChatGPT as a healthy competitor, but argued it still lacks real-time knowledge and memory. If you ask Alexa who scored in a specific World Cup game, it knows; ChatGPT does not. He also pointed out that Alexa already remembers personal details, like your favorite music.

Still, Amazon's own employees could not resist the hype. According to reporting from Insider, staff turned to internal Slack channels asking how to use ChatGPT for work, with some even using it as a coding assistant to polish internal code.

That prompted a company lawyer to warn employees against sharing confidential data, noting instances where responses looked uncomfortably close to Amazon internal assets. Interestingly, the lawyer also told staff that Amazon is broadly developing its own ChatGPT-like tech.

Why it matters

Jassy's comments show a tech giant forced into defense mode. While "we've been working on this for years" is the standard big-tech refrain, public momentum has undeniably belonged to Microsoft.

Upgrading Alexa into a full-fledged generative conversationalist sounds appealing, but the economics are tricky. Running LLM inference on millions of free smart speakers to answer simple trivia could become a massive money pit unless Amazon figures out new monetization.

The real battlefield right now is AWS. By positioning itself as the neutral home for startups like Stability AI, Amazon is trying to make sure it powers the rest of the ecosystem even if it does not own the consumer poster child. It is a smart defensive play, but developers will want to see actual models, not just cloud contracts.