Chatbots will change the way we shop
Imagine a world where you have a personal shopper at your disposal 24/7, an expert who can instantly recommend a gift to a friend or relative, even the hardest to buy, or scour the web to compile a list of the best libraries available within your tight budget. Better yet, they can analyze a kitchen appliance’s strengths and weaknesses, compare it to its seemingly identical competitors, and find you the best deal. Then, once you are happy with their suggestion, they will also take care of the purchase and delivery details.
But this ultra-informed buyer is not an informed human at all: he is a chatbot. This is not a far-off prediction either. Salesforce recently said it predicts that AI will drive $263 billion in online purchases this holiday season. This represents approximately 21% of all orders. And experts are betting that AI-enhanced shopping will become even more important in the coming years. By 2030, between $3 trillion and $5 trillion per year will come from agent commerce, according to research from the consulting firm McKinsey.
Unsurprisingly, AI companies are already investing heavily in making shopping through their platforms as seamless as possible. Google Gemini app can now exploit the powerful potential of the company Purchase Chart dataset on products and sellers, and can even use its agent technology to call stores on your behalf. Meanwhile, in November, OpenAI announced a ChatGPT Shopping Feature capable of quickly compiling shopping guides, and the company has deals with Walmart, Target, and Etsy to allow shoppers to purchase products directly within chatbot interactions.
Expect many more such types of deals to be made over the next year, as the time consumers spend chatting with AI continues to rise and web traffic from search engines and social media continues to fall.
—Rhiannon Williams
An LLM will make an important new discovery
I’m going to cover myself here, right off the bat. It’s no secret that great language models spew a lot of nonsense. Unless they have the luck of monkeys and typewriters, LLMs won’t discover anything for themselves. But LLMs still have the potential to expand the boundaries of human knowledge.
We got a glimpse of how this might work back in May, when Google DeepMind unveiled AlphaEvolve, a system that used the company’s Gemini LLM to propose new algorithms to solve unsolved problems. The breakthrough was combining Gemini with an evolving algorithm that checked its suggestions, selected the best ones and fed them back into the LLM to make them even better.
Google DeepMind used AlphaEvolve to come up with more efficient ways to manage the power consumption of Google’s data centers and TPU chips. These discoveries are important but do not change the situation. Again. Google DeepMind researchers are now pushing their approach to see how far it will go.
And others were quick to follow their example. A week after the release of AlphaEvolve, Asankhaya Sharma, an AI engineer in Singapore, shared OpenEvolve, an open source version of Google’s tool DeepMind. In September, Japanese company Sakana AI released a version of the software called SinkaEvolve. And in November, a team of American and Chinese researchers revealed AlphaResearch, which they say improves on one of AlphaEvolve’s already better-than-human math solutions.