Discover more from AI in Business
Tech Giants Engage In Battle for Generative AI Turf
New entrants include Amazon, Google, Elon Musk, Alibaba, Baidu and SenseTime; none are willing to cede turf to the early entrant, OpenAI
By John P. Desmond, Editor, AI in Business
The release of ChatGPT by OpenAI has set off a generative AI battle among the tech giants.
Sitting on the sidelines is not an option in this battle for AI and the future of software turf.
Among the developments:
Amazon has introduced Bedrock as a cloud service developers can use to generate text in a manner similar to that of ChatGPT, from OpenAI/Microsoft;
Google is planning a new AI search engine;
Elon Musk is starting another company, called X.ai;
Alibaba is releasing a generative AI model, and the Chinese government will require it to be true to the country’s norms.
Amazon’s Bedrock generative AI services will offer access to its own language model, called Titan, and access to language models from startups AI21 and Anthropic, a Google-backed startup, as well as a model for turning text into images from Stability AI, according to a recent account from CNBC. The Titan models are each designed to deliver specific services, such as to generate text for blog posts, emails or other documents, and another to help with search and personalization.
“Most companies want to use these large language models, but the really good ones take billions of dollars to train and many years, and most companies don’t want to go through that,” stated Amazon CEO Andy Jassy. “So what they want to do is they want to work off of a foundational model that’s big and great already and then have the ability to customize it for their own purposes. And that’s what Bedrock is.”
Generative AI products including ChatGPT have been guilty of “hallucination,” referring to results that do not match any data the model has been trained on, and thus, can be inaccurate or completely made up. An AWS VP, Bratin Saha, told CNBC that Amazon is “really concerned about accuracy and ensuring its Titan models produce high-quality responses.”
Customers will be able to customize Titan models with their own data, but that data will not be used to train the Titan models, according to Amazon. Bedrock is in a limited preview phase currently; pricing was not announced.
Google Hard at Work on Next-Gen Search
Google, meanwhile, is hard at work to get its next-generation search engine to market to protect its search business, according to a recent account in The New York Times.
Google in 2022 generated some 80 percent of its revenue, or $224 billion, from the search business, according to a recent account from Statista. The company is responsible for more than 84 percent of global desktop search traffic; in many markets, Google has no competition.
Under the project name Magi, Google’s new search features are being created by designers, engineers and executives working in a sprint fashion to develop the latest version, which is projected to offer more personalized services than the current experience, the Times reported.
Google has not yet announced the new features; no timetable for the release has been set. Google has recently unveiled the Bard chatbot, it’s first answer to ChatGPT, in a limited release phase.
An AI research leader for many years, Google’s effort is exemplified by its DeepMind lab in London, considered a pioneering AI research center that has used large language models to improve the quality of search results. But Google has held off on fully deploying its AI large language models, for fear they may generate false and biased statements. But now the game has changed; Google is on defense and needs to field a market response.
The Times reported that Google has more than 160 developers working full time on its next-generation search offering.
Google and AI got a little more mainstream on Sunday, April 16, when Sundar Pichai, Google’s CEO, was featured on 60 Minutes in an interview with correspondent Scott Pelley. The segment included a demo of Bard, which made an impression on Pelley. Google executive James Manyika was introduced, described as holding a new position at Google, with the responsibility to think about how AI and humanity can best coexist.
“AI has the potential to change many ways in which we’ve thought about society, about what we’re able to do, the problems we can solve,” Manyika stated on the segment, according to an account from CBSNews.
Alibaba, Baidu, SenseTime Active in China
Meanwhile in China, Alibaba Group Holding Lt. on April 11 announced it will release its own generative AI model “in the near future,” saying it would be integrated into the company’s apps, first the DingTalk workplace messaging app and later the Tmall Genie voice assistant, according to an account from Fox Business.
The new product will be called Tongyi Qianwen, which translates in English as “seeking an answer by asking a thousand questions.” Alibabi plans to open up Tongyi Qianwen to clients who want to build their own customized large language models.
To keep the AI generative model market in line in China, the Cyberspace Administration of China will require its developers to ensure that the AI-generated content is “accurate” and does not “endanger security.” Guidelines issued by the authority said the model’s content should “reflect the core values of socialism and must not contain subversion of state power.” Alibaba will need to submit the software for a security assessment prior to its release, according to the account.
Chinese tech company Baidu unveiled its generative AI model, Ernie Bot, in late March, positioning squarely against ChatGPT.
In addition, Chinese AI firm SenseTime on April 10 announced it would be releasing SenseChat, based on the company’s SenseNova model, which it has been developing for five years, according to an account from Reuters.
“We have been focusing on developing our big models. In addition to our powerful AI infrastructure … SenseTime has also developed the ability to deploy big models across our product line,” stated Xiaogang Wang, company cofounder. More specific plans for the rollout were not announced.
The US in 2019 put SenseTime on a trade blacklist after accusing the company of developing facial recognition programs that enable the surveillance of ethnic Uyghurs in the Xinjiang region by Chinese authorities. SenseTime opposed the ban and states it is working with authorities in an effort to resolve the situation.
Lastly for now, Elon Musk has created a new company, X.AI Corp., intended to be competition for OpenAI, according to an account in The Wall Street Journal. Musk was a cofounder of OpenAI; he left the company in 2018; since then, OpenAI has taken major investments from Microsoft. Musk has complained that ChatGPT is politically biased, and he seeks to create AI models that are more “truth-seeking,” without publicly stating what he means by that, according to the Journal account.
Musk has been recruiting for his new company. He recently hired Igor Babuschkin, a scientist with the DeepMind lab owned by Google/Alphabet, to head up X.AI Corp., the Journal reported.
(Write to the editor here; tell him what AI topic you want to read about in this newsletter.)
Thanks for reading AI in Business! Subscribe for free to receive new posts and support my work.