Picture credits: SOPA Images/Contributor/Getty Images
Google is rushing to participate in the sudden fervor for conversational AI, driven by the ubiquitous success of rival OpenAI’s ChatGPT. Bard, the company’s new AI experience, aims to “combine the breadth of global knowledge with the power, intelligence, and creativity of our great language models.” Not short of ambition, Google!
The model, or service, or AI chatbot, as you wish to describe it, was announced in a blog post by CEO Sundar Pichai. He emphatically notes Google’s refocusing around AI a few years ago, as well as the fact that the most influential concept (the Transformer) was created by the company’s researchers in 2017.
“It’s a really exciting time to be working on these technologies as we translate extensive research and breakthroughs into products that really help people,” Pichai writes. It’s hard not to wonder while reading this how Google managed to get so decisively overtaken by OpenAI, the latter now synonymous with the technologies the former pioneered.
The short explanation is that technology moves fast and big companies move slowly, and while Google released paper after paper and tried to figure out how to incorporate AI into its existing business strategies, OpenAI focused on creating of the best models and let people understand their own applications.
Bard shows Google taking a page from that playbook, releasing a “light” version of the model for testing. The model uses Google’s own LaMDA (Language Model for Dialogue Applications) to power a conversational AI that can also leverage information from the web. How exactly that isn’t apparent from the blog post, but it at least seems to be staying more or less up to date.
Bard “help[s] explain new discoveries from NASA’s James Webb Space Telescope to a 9-year-old, or learn about the best strikers in football right now, then get drills to develop your skills.
Google of course keeps the most up-to-date record of web content on Earth, and Bard will no doubt use this information to its advantage, but exactly how it processes and aggregates this information for you and your 9-year-old will that clear once people start using it.
The post notes that you can also use Bard to “plan a friend’s baby shower,” “compare two Oscar-nominated movies,” and “plan a trip to Ecuador.” One can imagine how an AI model could do any of these things using the various search results and firehoses of data that Google has access to, but that experiment will probably just tell you things, without do deep integrations with things like your calendar or airlines.
Of course, every conversational AI has to deal with the inevitable (nowadays, almost instantaneous) attempts to trick it into saying something hateful, stupid, or embarrassing. Google will surely record conversations with users “to ensure that Bard’s responses achieve a high level of quality, security, and grounding in real-world information.” The latter is clearly a jab through the OpenAI arcs, as well as Microsoft’s, since the former’s models don’t cite their sources and the latter’s short-lived Galactica invented them.
(Update: “In light of recent announcements”, Microsoft has now made public a previously confidential event to be held tomorrow in Redmond. The topic isn’t officially declared, but it’s widely expected to be a Bing-OpenAI tie-in that brings a next-gen language model to Microsoft’s perpetually beleaguered search engine. An early version of the features was reportedly tested and leaked by student Owen Lin, but we couldn’t confirm anything from this post.)
AI will be coming more directly to Google Search in the form of several new features “that can help synthesize information for questions where there is no right answer. Soon you will see these AI-powered features in research that distill complex information and multiple viewpoints into easy-to-digest formats, so you can quickly understand the big picture and learn more from the web,” the company said in a separate statement. -mail Nuance but format with bullets, I understood.
Although people will no doubt ask him for variations of the cart problem, the example provided is someone asking “Is piano or guitar easier to learn and how much practice does each have? he need?”
Not an unethical query (for the most part) but not necessarily with a simple result. But if among a hundred articles comparing different instrument learning rates there is some sort of consensus on how hard it is, with various equally common caveats and tips, Google can just suck them up and display them at the top of the results. of research.
The questions abound: isn’t this plagiarism? Will Sponsored Placements go above or below, and will they be included and/or promoted as part of the AI? What is considered a question with no good answer? Can users customize the results or the exploration process?
We could very well learn the answers to these questions at Google’s search and artificial intelligence event on Wednesday morning, an event that, strangely, isn’t mentioned in Pichai’s post. You can watch the live stream here at 6:30 a.m. PT, or check the front page for more information.