In the wake of ChatGPT’s viral success, DeepMind is preparing to unveil Sparrow: its own AI chatbot. A private beta could begin in the course of 2023. Find out everything you need to know about this tool, billed as more ethical than OpenAI’s…
Acquired by Google in 2014, DeepMind has made its mark in recent years with its many discoveries in the field of AI. Its researchers have notably created AlphaCode capable of coding better than humans, achieved a crucial breakthrough in drug creation with AlphaFold, and triumphed over the best Starcraft players with AlphaStar.
However, this pioneer of artificial intelligence has recently had the rug pulled out from under him. In 2022, its rival OpenAI made a splash with the Text-to-Image AI DALL-E 2, and more recently with the super-powered chatbot ChatGPT.
Given the phenomenal success of this tool, Demis Hassabis, CEO of DeepMind revealed to Time that the firm plans to deploy its own chatbot Sparrow in private beta sometime in 2023.
A more ethical chatbot than ChatGPT
At the end of last year, Sparrow was presented through a scientific study. It is described as ” a useful dialogue agent capable of reduce the risk of inappropriate responses “.
Indeed, DeepMind believes that chatbots like ChatGPT present potential dangers such as. “ imprecise or invented information “. This is why Sparrow has not yet been launched.
According to Demis Hassabis, the company wanted to ensure that its chatbot had important features that ChatGPT lacks. In particular, he believes it is essential that these AIs cite their sources.
If you’ve ever used ChatGPT, you’ll know that this AI is capable of feign intelligence on many subjects. However, a notion of morality and the ability to cite sources are indispensable.
However, to develop better rules for its AI, DeepMind needs input from experts on many subjects, as well as participation from the general public. This will be the main interest of the beta.
The article published by DeepMind also suggests that Sparrow will initially be more restricted than ChatGPT. ChatGPT has raised many concerns, notably because of its ability to generate malware or racist texts.
To avoid this danger, Sparrow will be subject to strict rules and will be able to refuse to answer questions if he feels it’s best to rely on humans.
In initial tests, DeepMind’s chatbot managed to provide a plausible, evidence-backed answer to 78% of factual questions.
Can DeepMind Sparrow save Google from ChatGPT?
Many experts consider ChatGPT a threat to the hegemony of search engines such as Google. In fact, Microsoft is planning to add this AI to Bing in order to finally compete with its great rival.
Of course, Google is well aware of this threat. As part of the Alphabet group, DeepMind could come to the rescue with Sparrow.
While OpenAI is preparing to launch the paid tool ChatGPT Pro and could deploy the GPT-4 model as early as this year, the firm is backed by another GAFAM: Microsoft, which is preparing to invest $10 billion and integrate this AI into all its products.
The AI war is launched between the tech giants, and could become their top priority. All we have to do now is wait a few months, to witness a live debate between Sparrow and ChatGPT…