Who is the competition to GPT: Meet the candidates

Today, ChatGPT is two months and one day old. Believe it or not, less than nine weeks ago, OpenAI launched what it simply described as an ‘early demo’ of the GPT-3.5 series. Quickly, ChatGPT captured the imagination, sparked feverish excitement, and for some, a slight fear. Since then, the capabilities of the tool, as well as its limitations and hidden dangers, have been well established, and all signs of slowing its development quickly disappeared when Microsoft announced its investment plans in OpenAI.

Can anyone catch up and compete with OpenAI and ChatGPT? It seems that every day competitors, both new and old, are entering the AI ring, and Reuters reports that the Chinese internet giant Baidu also plans to launch an AI chatbot service similar to OpenAI’s ChatGPT in March.

And while we wait for the Chinese version, we present those who are already here and who could potentially challenge ChatGPT.

Anthropic – Claude

The San Francisco startup Anthropic is already valued at around $5 billion, founded in 2021 by several researchers who left OpenAI. Anthropic attracted more attention last April when, after less than a year of existence, it suddenly announced an incredible $580 million in funding. However, it turned out that the vast majority of those millions came from Sam Bankman-Fried and people from FTX, the cryptocurrency platform that is now bankrupt and accused of fraud. Anthropic has developed an AI chatbot, Claude — available in closed beta through Slack integration — which reports say is similar to ChatGPT. Anthropic is working on ‘building interpretable and manageable artificial intelligence systems’, as they describe themselves, and the chatbot Claude was created through a process called ‘Constitutional AI’ which they say is based on concepts such as benevolence, non-harmfulness, and autonomy.

Google – LaMDA

You may remember LaMDA when Blake Lemoine, a Google engineer, was fired for his claims that LaMDA (Language Model for Dialogue Applications) is sentient. More specifically, he said he believes LaMDA is a person, which automatically got him fired, and he probably got off easy as he could have ended up in worse places. Although it is certainly not a person, LaMDA is still considered one of the biggest competitors to ChatGPT. Launched in 2021, Google stated in a blog post that LaMDA’s conversational skills ‘have been developing for years’. Like ChatGPT, LaMDA is built on the Transformer, a neural network architecture invented and opened by Google Research in 2017. Just like ChatGPT, LaMDA is trained for dialogue. An article in the New York Times from January 20 states that last month, Google founders Larry Page and Sergey Brin met with company executives to discuss ChatGPT, which could pose a threat to Google’s $149 billion search business.

Character AI

What happens when engineers who developed Google’s AI LaMDA get fed up with big tech bureaucracy and decide to venture into business challenges on their own? Well, just three months ago, Noam Shazeer (who was also one of the authors of the original Transformer paper) and Daniel De Freitas launched Character AI, their new AI chatbot technology that allows users to chat and role-play with, well, anyone, living or dead — the tool can mimic historical figures like Queen Elizabeth and William Shakespeare, for example, or fictional characters like Draco Malfoy. The startup has stated that they aim to raise as much as $250 million in a new funding round, which is a huge sum for a startup with a product that is still in beta. Currently, the technology is reported to be free to use, and in October, Shazeer and De Freitas told the Washington Post that they left Google to ‘bring this technology to as many hands as possible’.

DeepMind – Sparrow

Recently, DeepMind’s CEO and co-founder Demis Hassabis told Time that DeepMind is considering releasing its chatbot Sparrow in a ‘private beta version’ sometime in 2023. He added that it is currently ‘wise to be cautious’ and that the company is currently working on enhancing learning-based features such as source attribution, which is something ChatGPT lacks. DeepMind, a subsidiary of Google’s parent company Alphabet, introduced Sparrow last September, announcing it as a step towards creating safer, less biased machine learning systems. DeepMind states that Sparrow is ‘a useful dialogue agent that reduces the risk of unsafe and inappropriate responses’. However, DeepMind has stated that it considers Sparrow a research-based model, with a proof of concept that is not ready for implementation, according to Geoffrey Irving, a safety researcher at DeepMind and the lead author of the paper presenting Sparrow. We have not implemented the system because we believe it has a lot of biases and shortcomings of other kinds, Irving told VentureBeat last September.