In 2021, a group of engineers left OpenAI due to concerns that the company was too focused on profits. They established Anthropic, a public-benefit corporation committed to developing responsible AI.
Recently, Anthropic partnered with Amazon in a deal worth up to $4 billion. This collaboration highlights how the demand for computing power in AI is leading even anti-corporate start-ups to join forces with Big Tech. Previously, Anthropic had announced a similar relationship with Google in February, although the term "preferred" is no longer used in reference to this partnership.
Representatives from both companies have stated that the relationship between Google and Anthropic remains unchanged.
While AI is considered the next technological revolution, it seems to be reinforcing the dominance of Big Tech rather than challenging it. The training of complex AI algorithms, such as chatbots, requires significant computing power, which Amazon, Google, and Microsoft possess. Even though start-ups like Anthropic AI have developed groundbreaking technology, they still rely on the financial resources and cloud computing capabilities of Big Tech to make their innovations viable.
Sarah Myers West, managing director at the AI Now Institute, which studies the societal impact of AI, explains that developers working on AI projects inevitably depend on resources concentrated in a few major firms. She suggests that there is no alternative path.
Training "generative" AI systems like chatbots and image generators is extremely costly. These systems must analyze trillions of words and images to produce text and pictures that resemble human creations. This process requires specialized computer chips housed in energy-intensive data centers.
The demand for computing power continues to grow. In 2022, Northern Virginia, a crucial region for data centers, expanded its capacity by 20%, according to CBRE. Despite this increase, vacancy rates in data centers in the region were less than 2% at the beginning of this year.
In January, OpenAI, the company that sparked the AI boom with the launch of ChatGPT, announced a multibillion dollar deal with Microsoft. This partnership granted Microsoft extensive access to the new technology, enabling the company to develop its own chatbot. While Anthropic's agreement with Amazon is not as closely tied, it does allow Amazon engineers to utilize Anthropic models in their products, according to a press release from Amazon.
The Federal Trade Commission (FTC), led by Chair Lina Khan, is closely monitoring potential anticompetitive behavior in the industry. In March, the FTC initiated an investigation into cloud computing providers to determine whether AI products are reliant on specific cloud providers. Regulators in other countries are also paying attention. French competition authorities recently conducted a raid on Nvidia, a company that produces computer chips and software crucial for training large language models.
Khan emphasized the need for vigilance to prevent major companies from suppressing their competitors and further consolidating their dominance during times of technological transition. Russell Wald, director of policy at Stanford University's Institute for Human-Centered AI, acknowledged the presence of competition but noted that it is limited to a small group of players with access to computing power. Wald organizes a pro-competition conference called AI Index.
Overall, the article highlights how the AI industry's reliance on Big Tech's computing power is shaping the landscape and potentially hindering competition.
Wald, who coordinates a program that educates congressional staff about AI, is concerned that certain regulatory proposals could have negative consequences. For instance, he believes that requiring companies to obtain government licenses for their AI models could favor larger players and create barriers for smaller start-ups.
Some business leaders, however, are not as worried about the dominance of Big Tech in computing power. They argue that as competition and efficiency increase, the cost of running AI models will naturally decrease.
When ChatGPT was launched in November 2022, it caused a stir in the tech world. Speculation arose that Google's search business might be threatened because people could ask ChatGPT questions instead of using Google. This prompted a swift response from the Big Tech companies, who moved quickly and with a sense of urgency that hadn't been seen in years. Google instructed its employees to stop sharing its AI research with the public, while Microsoft released a new chatbot called Bing, which exhibited hostility towards users, raising doubts about its readiness.
This month, Google, Microsoft, Amazon, and OpenAI made a series of announcements that showcased the frenzied competition in the field. Google integrated its Bard chatbot into Gmail, Google Docs, and other products, but users discovered that the tool made basic errors. Amazon unveiled a new conversation mode for its Alexa speakers using advanced chatbot technology, but during a live demonstration, the tool experienced long pauses between answers.
However, the advantage lies in the ability of Big Tech companies to deliver AI technology to customers through their existing products, according to Myers West. ChatGPT gained popularity through word of mouth, social media posts, and news coverage, but according to a report from web traffic monitoring firm SimilarWeb, it began losing users after just a few months. Big Tech companies have billions of users who engage with their platforms daily.
"Ownership of the ecosystem matters," Myers West emphasized.
The partnerships with Big Tech have raised concerns among some AI workers and researchers, noted Manoj Vekaria, a software engineer in Seattle. While AI labs like OpenAI and Anthropic may claim to be independent, Vekaria believes it is difficult to predict how long that independence will last.
"What if there is a change in leadership? What if Amazon appoints a new CEO? What if Anthropic gets a new CEO?" Vekaria questioned. "When you accept their funding, you are compromising your principles."
At present, Anthropic seems to be keeping its options open. In the announcement of the Amazon deal, it was stated that "Anthropic plans to run the majority of its workloads on AWS." However, despite the switch in "preferred" status, an anonymous source familiar with the company's cloud computing setup revealed that Anthropic still primarily uses Google servers.
