The Linkielist

Linking ideas with the world

The Linkielist

Daily Archives: May 2, 2020

Facebook releases Blender AI Chatbot sources

Posted on by

  • Facebook AI has built and open-sourced Blender, the largest-ever open-domain chatbot. It outperforms others in terms of engagement and also feels more human, according to human evaluators.

  • The culmination of years of research in conversational AI, this is the first chatbot to blend a diverse set of conversational skills — including empathy, knowledge, and personality — together in one system.

  • We achieved this milestone through a new chatbot recipe that includes improved decoding techniques, novel blending of skills, and a model with 9.4 billion parameters, which is 3.6x more than the largest existing system.

  • Today we’re releasing the complete model, code, and evaluation set-up, so that other AI researchers will be able to reproduce this work and continue to advance conversational AI research.

[…]

As the culmination of years of our research, we’re announcing that we’ve built and open-sourced Blender, the largest-ever open-domain chatbot. It outperforms others in terms of engagement and also feels more human, according to human evaluators. This is the first time a chatbot has learned to blend several conversational skills — including the ability to assume a persona, discuss nearly any topic, and show empathy — in natural, 14-turn conversation flows. Today we’re sharing new details of the key ingredients that we used to create our new chatbot.

Some of the best current systems have made progress by training high-capacity neural models with millions or billions of parameters using huge text corpora sourced from the web. Our new recipe incorporates not just large-scale neural models, with up to 9.4 billion parameters — or 3.6x more than the largest existing system — but also equally important techniques for blending skills and detailed generation.

[…]

We’re currently exploring ways to further improve the conversational quality of our models in longer conversations with new architectures and different loss functions. We’re also focused on building stronger classifiers to filter out harmful language in dialogues. And we’ve seen preliminary success in studies to help mitigate gender bias in chatbots.

True progress in the field depends on reproducibility — the opportunity to build upon the best technology possible. We believe that releasing models is essential to enable full, reliable insights into their capabilities. That’s why we’ve made our state of the art open-domain chatbot publicly available through our dialogue research platform ParlAI. By open-sourcing code for fine-tuning and conducting automatic and human evaluations, we hope that the AI research community can build on this work and collectively push conversational AI forward.

 

Read the paper here.

 

Get the code here.

Source: A state-of-the-art open source chatbot

AI can’t be legally credited as an inventor, says US Patent Office

Posted on by

Artificial intelligence has myriad use cases, but it turns out inventing devices isn’t one of them — at least in the eyes of the US Patent and Trademark Office. The agency issued a decision on two patent applications for devices created by an AI system, determining that only humans can legally be credited as inventors.

The items in question — an emergency flashlight and a shape-shifting drink container — were the brainchildren of a system called DABUS. The Artificial Inventor Project filed the applications last year on behalf of the AI’s creator, Stephen Thaler. AIP lawyers argued that, since Thaler didn’t have any expertise in either of those types of products and couldn’t have come up with them by himself, DABUS should be the credited inventor.

The USPTO wasn’t buying it. The agency noted that US patent law uses pronouns and language such as “whoever” to refer to inventors. It wrote that “only natural persons may be named as an inventor in a patent application” as the law stands. The UK Intellectual Property Office and the European Patent Office previously rejected AIP applications on similar grounds, despite their belief that the devices were patent-worthy.

There was no suggestion, however, that DABUS itself might obtain any patents. Thaler himself was the applicant. “Machines should not own patents,” the AIP says on its website. “They do not have legal personality or independent rights, and cannot own property.”

Source: AI can’t be legally credited as an inventor, says USPTO | Engadget

However, patent rights can extend to the relatives of a dead person? I’m pretty sure dead people have no legal personality and can’t own property either

Virgin Galactic’s spaceship flies from its new home base for the first time

Posted on by

The pieces are finally starting to come together for Virgin Galactic’s space tourism — the company has flown SpaceShipTwo from Spaceport America for the first time. It was just a glide test from 50,000 feet up, but the flight let the spaceport fulfill its intended purpose and gave pilots familiarity with the New Mexico airspace. This will also help Virgin compare performance against similar maneuvers from earlier tests.

And before you ask: yes, Virgin took steps to keep crews safe during the COVID-19 pandemic. It reworked operational elements to keep people apart, and required “universal” mask usage.

There are still more test flights in the pipeline. Even so, this nudges Virgin considerably closer to its goal of taking paying passengers into space. The company is certainly under pressure to get things up and running quickly. Its financial situation has been rough for a while, and it won’t turn around until customers get what they’re paying for.

Source: Virgin Galactic’s spaceship flies from its new home base for the first time | Engadget

Trolls, bots flooding social media with anti-quarantine disinformation

Posted on by

Christopher Bouzy, the founder of bot tracking platform Bot Sentinel, conducted a Twitter analysis for Business Insider and found bots and trolls are using hashtags like #ReOpenNC, #ReopenAmericaNow, #StopTheMadness, #ENDTHESHUTDOWN, and #OperationGridlock to spread disinformation. According to Bouzy, the bots and trolls are spreading conspiracy theories about Democrats wanting to hurt the economy to make Trump look bad, Democrats trying to take away people’s civil liberties, and Democrats trying to prevent people from voting. The accounts are also using false data to underplay the threat of the coronavirus.

[…]

“Inauthentic accounts are amplifying disinformation and inaccurate statistics and sharing false information as a reason to reopen the country,” Bouzy says. “Many of these accounts are also spreading bizarre conspiracy theories about Democrats using COVID-19 as a way to take away American freedoms and prevent Americans from voting.”

[…]

“Inauthentic accounts are downplaying the seriousness of COVID-19, and they sharing inaccurate information about the mortality rate of the virus. The problem is significant because many of these inauthentic accounts are retweeted by other larger accounts, which increases their reach and visibility.”

According to the New York Times, Chinese operatives spread claims on social media in mid-March that the Trump administration was going to lock down the entire country and enforce this lockdown with soldiers on the streets. The White House’s National Security Council later tweeted that these claims were false. That was just some of the disinformation that’s been spread on social media by inauthentic sources.

[…]

Brooke Binkowski, managing editor of the fact-checking website Truth or Fiction and former managing editor of Snopes, tells Business Insider that the media has been struggling with its coverage of the protests, which she says are “completely inauthentic and coordinated.”

“Journalists are largely missing that fact in their bids to find ‘other sides to the story,'” Binkowski says.

[…]

She believes that the disinformation is being spread by trolls and bots but also by “useful idiots.”

“Empowering violent extremists is a very old method for collapsing unstable states,” Binkowski says. “This is the end result of weaponized disinformation — it’s doing its job. It would have been the virus or it would have been something like a fire, or a hurricane, or an earthquake. But disinformation purveyors are nothing if not opportunistic.”

Source: Trolls, bots flooding social media with anti-quarantine disinformation – Business Insider