Listen to this story
|
Looks like OpenAI needs journalists to make GPT-4 better. In a surprising move, the AI research company recently announced a partnership with German mediahouse Axel Springer to integrate the latest news from publications such as Politico and Business Insider in ChatGPT.
In turn, the partnership brings financial benefits for Axel Springer. OpenAI will compensate the news publication for using the data obtained from them to train its models. This is not the first time OpenAI has partnered with the media; earlier this year, it collaborated with the Associated Press.
Why Germany?
OpenAI does not share a good rapport with the European Union. Back in April, Germany had considered banning ChatGPT due to privacy concerns. Now, OpenAI partnering with a German mediahouse might provide them with some respite from government authorities.
In recent events, the EU has passed the EU AI Act, which stipulates that foundation models must comply with specific transparency obligations before being placed on the market. The EU differentiates general-purpose AI models based on risks, and OpenAI falls under the high-risk category. Axel Springer’s partnership would assist OpenAI in demonstrating to EU officials that it has not used stolen data to train its models.
Interestingly, earlier this year, OpenAI lobbied the EU and argued with European officials not to consider its general-purpose AI systems — including GPT-3, the precursor to ChatGPT, and the image generator Dall-E 2 — as ‘high risk’.
Moreover, during the EU discussion, Germany, France, and Italy were not in favour of imposing the act on open-source companies, and subsequently, they were exempted. Notably, Germany’s AI startup, Aleph Alpha, recently secured an impressive $500 million in a Series B round which is considered one of the European OpenAI competitors.
OpenAI’s Challenge to Grok
Interestingly, this partnership comes at a time when everyone is praising Grok for its capability to have access to real time information through X. Musk recently rolled out Grok to Premium+ subscribers on X, and in a further expansion, xAI is extending Grok to users in India and 46 other countries, including Australia, Canada, Malaysia, New Zealand, and Singapore.
It’s very likely that OpenAI will provide access to real time news only to ChatGPT Plus users, similar to Musk’s approach with Grok. Interestingly, OpenAI has now re-enabled ChatGPT subscriptions which were paused after OpenAI’s DevDay due to high surge in traffic.
At the moment, X boasts a larger paid user base compared to ChatGPT Plus. X has roughly around 640K premium users, while, on the other hand, ChatGPT has 180 million users. However, there are no numbers available to comment on how many of them are paid users.
Grok has a huge advantage that it is not dependent on any one particular media house as such and has access to every information users post on it. Recently citizen journalism has gained immense popularity among the masses and Musk has also advocated for it.
While OpenAI is trained on data available on the internet, offering news from selected media houses could introduce bias into the responses it generates, raising concerns for OpenAI’s safety team.
Despite the drawbacks, with this partnership, ChatGPT might improve in generating accurate information and reduce hallucinations. On X, there is a vast amount of information, but discerning what is correct from what is false can be a tedious task.
A few days ago, a very interesting incident occurred where it was revealed that Grok is also accidentally trained on OpenAI’s data.
A user on X shared a screenshot where Grok’s response was based on OpenAI. However, xAI’s engineer, Igor Babuschkin, soon clarified the situation, and said, “The issue here is that the web is full of ChatGPT outputs, so we accidentally picked up some of them when we trained Grok on a large amount of web data.”
Though Grok currently has an advantage over OpenAI in providing real-time information, this can be considered OpenAI’s attempt to get back at them.