Hitting the Books: How much that insurance monitoring discount might really be costing you

Machine learning systems have for years now been besting their human counterparts at everything from Go and Jeopardy! to drug discovery and cancer detection. With all the advances that the field has made, it’s not unheard of for people to be wary of ro…

Meta unleashes BlenderBot 3 upon the internet, its most competent chat AI to date

More than half a decade after Microsoft’s truly monumental Taye debacle, the incident still stands as stark reminder of how quickly an AI can be corrupted after exposure to the internet’s potent toxicity and a warning against building bots without sufficiently robust behavioral tethers. On Friday, Meta’s AI Research division will see if its latest iteration of Blenderbot AI can stand up to the horrors of the interwebs with the public demo release of its 175 billion-parameter Blenderbot 3.

A major obstacle currently facing chatbot technology (as well as the natural language processing algorithms that drive them) is one of sourcing. Traditionally, chatbots are trained in highly-curated environments — because otherwise you invariably get a Taye — but that winds up limiting the subjects that it can discuss to those specific ones available in the lab. Conversely, you can have the chatbot pull information from the internet to have access to a broad swath of subjects but could, and probably will, go full Nazi at some point. 

“Researchers can’t possibly predict or simulate every conversational scenario in research settings alone,” Meta AI researchers wrote in a Friday blog post. “The AI field is still far from truly intelligent AI systems that can understand, engage, and chat with us like other humans can. In order to build models that are more adaptable to real-world environments, chatbots need to learn from a diverse, wide-ranging perspective with people ‘in the wild.'” 

Meta has been working to address the issue since it first introduced the BlenderBot 1 chat app in 2020. Initially little more than an open-source NLP experiment, by the following year, BlenderBot 2 had learned both to remember information it had discussed in previous conversations and how to search the internet for additional details on a given subject. BlenderBot 3 takes those capabilities a step further by not just evaluating the data it pulls from the web but also the people it speaks with.  

When a user logs an unsatisfactory response from the system— currently hovering around 0.16 percent of all training responses — Meta works the feedback from the user back into the model to avoid it repeating the mistake. The system also employs the Director algorithm which first generates a response using training data, then runs the response through a classifier to check if it fits within a user feedback-defined scale of right and wrong. 

“To generate a sentence, the language modeling and classifier mechanisms must agree,” the team wrote. “Using data that indicates good and bad responses, we can train the classifier to penalize low-quality, toxic, contradictory, or repetitive statements, and statements that are generally unhelpful.” The system also employs a separate user-weighting algorithm to detect unreliable or ill-intentioned responses from the human conversationalist — essentially teaching the system to not trust what that person has to say. 

“Our live, interactive, public demo enables BlenderBot 3 to learn from organic interactions with all kinds of people,” the team wrote. “We encourage adults in the United States to try the demo, conduct natural conversations about topics of interest, and share their responses to help advance research.”

BB3 is expected to speak more naturally and conversationally than its predecessor, in part, thanks to its massively upgraded OPT-175B language model, which stands nearly 60 times larger than BB2’s model. “We found that, compared with BlenderBot 2, BlenderBot 3 provides a 31 percent improvement in overall rating on conversational tasks, as evaluated by human judgments,” the team said. “It is also judged to be twice as knowledgeable, while being factually incorrect 47 percent less of the time. Compared with GPT3, on topical questions it is found to be more up-to-date 82 percent of the time and more specific 76 percent of the time.”

Hitting the Books: How Moderna dialed-in its vaccine to fight COVID’s variants

The national news cycle may have largely moved on from coverage of the COVID-19 pandemic — despite, as of this writing, infections being on the rise and more than 300 deaths tallied daily from the disease. But that certainly doesn’t diminish the unprec…

The best trail cameras for keeping an eye on your backyard wildlife

Daffodils flourishing in sidewalk cracks, pigeons and starlings congregating on overhead power lines, rats living in your apartment walls — no matter how urban humans strive to make our environments, nature’s flora and fauna will make themselves right …

GM’s ‘EV Live’ online showroom is here to answer your most pressing EV questions

It’s been a while since EVs shed their fringe curiosity reputation and become a mainstream transportation technology, but they’re still not yet ubiquitous enough that the general public is really comfortable with the vagaries of their day to day use. Basically, EVs are the shiny new toy and people still have questions. GM is here to answer them. The company announced on Monday that it is opening an online showroom/studio, dubbed EV Live, that will host Q&A sessions with the general public about electrification, the ins and outs of EV ownership and GM’s Ultium 360 charging network and electric vehicle offerings.

The free service will allow anyone in the US, over the age of 18, with an internet connection to contact one of GM’s EV liaisons to “answer EV-related questions in real-time and give virtual tours of the EV Live studio.” That studio will feature mockups of GM’s home and public chargers, the company’s battery technology, and of course GM EVs. The liaisons will be able to speak on a wide range of subjects — from the engineering and chemistry that goes into the batteries, to explaining the home charger installation process and select a certified vendor — but don’t expect the answers to be all-encompassing.     

“If somebody’s got a question about a Tesla battery pack, I’m sure they’ve done a lot of resources at their fingertips,” Hoss Hassani, GM vice president of EV Ecosystem, said during a press call on Friday. “We want to talk to people about EV considerations overall where the opportunity presents to talk specifically about the GM advantage.”

“We are not looking for our EV specialists to offer any editorial commentary, or get into a political discussion about federal policy, or state policy, or any of that,” he added. The showroom is focused primarily on electric cars, trucks and SUVs but Hassani hinted that ebikes, electric ATVs and other offroad electric transports could eventually become topics of discussion as well. 

GM expects both prospective EV buyers and recent purchasers to find value in this service. “If you’re someone who owns an EV, if you drove off a lot and then realized — like many of us do — ‘oh shoot, I have a whole bunch of questions that I didn’t get answered,’ this is an awesome place to come to to understand how you can make the most of what you’re already driving,” a GM representative noted during the call.  

Visitors will be able to schedule a live one-on-one tour with a liaison — on-demand live group tours and prerecorded walkthroughs are coming later this year — and ask questions either through voice or text chat. But before you go whipping out your junk on camera, know that the liaisons will not abide. 

“The staff are empowered,” Hassani said. “If they find a conversation is just headed in a direction that is untoward, or that somebody is treating them inappropriately… well, it’s very easy to disconnect the call.” The studio will be open Monday to Thursday from 9am to midnight ET, Friday from 9am to 9pm ET, and Saturday to Sunday, 11am to 7pm ET.