OpenAI's Losses Double to $540 Million Amid ChatGPT Expense
OpenAI, the AI research company known for its AI programs like ChatGPT and image generator Dall-E 2, reportedly doubled its losses to $540 million in 2022 due to development expenses for the ChatGPT. In addition to investing in cloud computing for powering its language models and training new versions, data costs are also expected to skyrocket as platforms like Reddit and StackOverflow introduce policies that charge AI firms access to their previously free data sets.
The company may now try to raise as much as $100 billion in the coming years to achieve its aim of developing artificial general intelligence (AGI), which is advanced enough to improve its own capabilities. OpenAI CEO Sam Altman has privately suggested the firm's monopoly on AI after its recent acquisition of the domain AI.com, redirecting it to ChatGPT, and filing a trademark on 'GPT.' OpenAI is also keen to acquire fresh data sets for its software, some of which are currently not on the internet.
Although revenue has picked up, reaching an annual pace of hundreds of millions of dollars weeks after the launch of a paid version in February, costs are likely to keep rising. This comes as more customers use AI technology, and the company trains future versions of the software. According to The Information report, OpenAI losses are due to employment costs, citing close sources with knowledge of the firm's financials.
OpenAI reportedly lured key employees like engineers and research experts from Google to work on its ChatGPT, which went viral immediately after its release, prompting other tech firms to join the bandwagon for chatbots or incorporate the tech into their services. Jacob Devlin, who was an AI researcher at Google, tried to warn Google not to use ChatGPT data to train its AI chatbot Bard as it violated OpenAI's terms of service. Additionally, Bard's answers resembled ChatGPT's.
However, these data sets are a point of contention elsewhere. Companies like Samsung are wary of their proprietary data slipping into the ChatGPT data pool, as evidenced by a recent ban on generative AI usage with company materials. Other experts believe that as the AI arms race intensifies, OpenAI's costs will escalate even further, likely deepening its losses.
Meta Launches AI Sandbox for Advertisers
Meta, formerly known as Facebook, announced its own AI Sandbox for advertisers to trial generative AI tools and services. The Sandbox gives advertisers a chance to experiment with "early versions of new tools and features, including generative AI-powered ad tools" for text variation, background generation, and image cropping. Advertising has long been a pillar of the Meta business model, so it's unsurprising that Zuckerberg and co are using AI to achieve better results in this department.
While some will claim Meta's embrace of AI is a new development, that isn't the case. Meta has "used machine learning and AI to power all of our apps and services, including our ads system" since the earliest days of News Feed in 2006. The AI Sandbox isn't out in the wild yet, as only a small core of Meta advertisers is tinkering with the tools. Meta has earmarked July for a wider roll-out of AI Sandbox to advertisers, with plans to add some of the features into products by the end of 2023.
The major advantage of AI is that it can save money and time while achieving better performance. If Meta can make advertisers' lives easier while charging them to advertise on its platforms, so much the better.
