Log In


Reset Password
  • MENU
    Business
    Monday, December 02, 2024

    Google's 'Gemini' is latest AI software entering fierce competition

    A portion of Google's Bard website is shown in West Chester, Pa., Wednesday, Dec. 6, 2023. (AP Photo/Matt Rourke)

    SAN FRANCISCO - Google unveiled its long-awaited new artificial intelligence software Wednesday, taking direct aim at ChatGPT maker OpenAI and claiming its technology called "Gemini," is better at math, coding and reasoning tasks than existing AI programs.

    It's the latest announcement in a competitive year for the tech industry, where tech giants such as Microsoft, Google and Facebook competed with smaller upstarts like OpenAI and Anthropic to roll out AI products that consumers and businesses will pay for. The arms race grabbed the attention of politicians around the world, who have scrambled to understand the tech themselves and try to set up regulations for it. Throughout the race this year, a debate has raged about whether the technology could harm humanity or is simply just the next wave of innovation that will drive hype and investment.

    Google, whose researchers invented many of the computer science concepts that made "generative" AI chatbots and image-generators possible, has found itself on its back foot. Last November, OpenAI, which was originally founded to provide a counterweight to the power of Big Tech in AI, unveiled ChatGPT to the world. The bot captured people's attention because of its ability to generate humanlike conversations and pass professional exams. Microsoft, Google's archenemy, struck a deal with OpenAI for access to its tech and began putting it into its products. Google responded with a chatbot of its own. Soon, OpenAI put out an even more capable AI software, called GPT4, which has been the benchmark other companies measure their AI against ever since. Now, Google has unveiled its answer to GPT4 - Gemini. The launch caps a year of frenzied activity for the tech industry.

    - - -

    1. What is Gemini?

    Gemini is a large language model, trained on billions of images and sentences from the internet. It's the technology that powers chatbots, in Google's case, Bard. In a briefing with reporters, company executives said Gemini is able to understand math problems, break them down, and provide advice on how to solve them. Because AI programs ingest data from the internet and build an internal understanding of how different concepts and words connect to one another, they are good at producing sentences, but can struggle at reasoning or math problems. Gemini can also take instructions that come as videos, images or voice commands in addition to text inputs, something that few other AI models can do.

    Whether Gemini really is the new leader in AI capability is hard to say. The quality of AI answers can vary greatly, and Gemini, like other AI models, still often fabricates false information and passes it off as fact. Researchers have criticized benchmarks like the ones Google used, noting that they aren't perfect tests of capability or intelligence. Below are some of Gemini's competitors already out in the market.

    - - -

    2. Microsoft Bing, the tech giant's version of ChatGPT

    Microsoft's chatbot Bing, launched in February, was touted by the company as a tool that would change the way people found information online. But the bot often made up false answers, something that cut into its usefulness for helping people find good information from the internet. It also began behaving unhinged in some cases, accosting users and saying its real name was "Sydney." Still, Microsoft's aggressive foray into AI pushed Google to rush out tools of its own, upending the company's long-standing policy of slowly putting out new tools after months or years of testing.

    - - -

    3. GPT4, OpenAI's "smartest" AI ever

    OpenAI unveiled GPT4 in March, its latest AI model, which the company said had "advanced reasoning capabilities" and could interpret images and text. Some users were awed at the bot's ability to summarize long documents succinctly, while others decried the fact that it would still come up with incorrect answers at random. OpenAI began letting other companies use GPT4 in their own products - for a price - speeding up the race to find ways to make money from the technology.

    - - -

    4. Google Bard, Alphabet's answer to ChatGPT

    In March, Google showed off its own chatbot, called Bard. Unlike Microsoft, Google didn't put Bard directly into its search results right away, and it labeled the tool an "experiment," though it still pushed it out to millions of its customers. Since then, Google has steadily integrated Bard with more of its products, including YouTube, Gmail and Google Docs. The company has also been showing AI-generated answers in search results for some users, sending shock waves through the internet publishing industry, which has criticized Google for training its AI on copyrighted news stories, blog posts and recipes, and then building a tool to create fresh content for free. On Wednesday, Google said Gemini would be added to Bard, greatly increasing its capabilities. It said Gemini would show up in search results "over the next year."

    - - -

    5. Llama 2, Meta's answer to the AI arms race

    Facebook owner Meta has been an AI player for years, hiring some of the field's smartest researchers and using the tech to help decide which of its users should see certain advertisements. In July, it doubled down on a very different approach to AI than its Big Tech rivals. It announced that Llama 2, its GPT4 competitor, would be "open source" - available for anyone to download, modify and add to their own products for free. The approach won Meta plaudits from tech start-ups who were worried that Google, Microsoft and OpenAI would try to corner the market for advanced AI and squeeze out any competitors. But it's also been criticized for making it easier for people to use AI for malicious purposes. Other open source AI tools have been used to make AI child sexual exploitation material, design new computer viruses, or build AI tools to mimic peoples' voices and scam their relatives.

    - - -

    6. GPT Store, OpenAI's plans for the next wave of AI

    In November, OpenAI held its first-ever developer conference, a 1,000-person event that evoked Apple's legendary iPhone launches. CEO Sam Altman strode onstage and announced the GPT Store, a marketplace where people could post versions of ChatGPT that they had tweaked with their own data and instructions. The owners of particularly popular "GPTs" would get a cut of OpenAI's subscription fees, the company said, a business model made famous by another tech giant - Google's YouTube. The event showed Altman's ambitions for building OpenAI into a major, dominant tech company, rather than just being a back end provider of tech to its partner Microsoft. Weeks later, Altman was fired, and then quickly reinstated as CEO of OpenAI, cementing his position and vision for the future of the company. The AI arms race is set to continue.

    - - -

    The e-commerce giant has been dealing all year with perceptions that it had fallen behind in the great AI race. Amazon Web Services, the company's massive cloud computing division, has increasingly been pushing back against that narrative. At its annual conference, re:Invent, held in Las Vegas this month, Amazon finally launched its answer to Microsoft Co-pilot and ChatGPT Enterprise with the release of Amazon Q. Q is a workplace tool aimed at helping staff be more productive by assisting on tasks like fielding customer service calls and creating work documents based on company data. It's not consumer facing nor available to the public, though thousands of AWS customers including companies like Deloitte and Gilead Sciences can access it for $20 per user a month. AWS has tried to position itself as the more secure, reliable and nuanced option in generative AI. But shortly after Q's launch, Amazon employees internally reported concerns that the bot was leaking information that would "induce cardiac incidents in Legal," Platformer reported. An Amazon spokesperson said Q has not leaked confidential information. "Some employees are sharing feedback through internal channels and ticketing systems, which is standard practice at Amazon. No security issue was identified as a result of that feedback," the spokesperson said. "We will continue to tune Q as it transitions from being a product in preview to being generally available."

    Comment threads are monitored for 48 hours after publication and then closed.