Tag Archives: Meta

Meta’s AI Chatbot Needs Some Work

Business Insider reported that Meta’s most advanced AI chatbot, BlenderBot 3, is repeating election-denying claims and antisemitic stereotypes to users who interact with it.

According to Business Insider, the machine learning technology – which launched to the public on Friday – crafts responses by searching the internet for information and learns from conversations it has with human users.

On August 5, 2022, Meta posted about BlenderBot 3. Part of the blog post included the following information:

“To improve BlenderBot 3’s ability to engage with people, we trained it with a large amount of publicly available language data. Many of the datasets used were collected by our own team, including one new dataset consisting of more than 20,000 conversations with people predicated on more than 1,000 topics of conversation. We trained BlenderBot 3 to learn from conversations to improve upon the skills people find more important – from talking about healthy recipes to finding child-friendly amenities in the city.

“When the chatbot’s response is unsatisfactory, we collect feedback on it. Using this data, we can improve the model so that it doesn’t repeat its mistakes.”

Meta also wrote: “We understand that not everyone who uses chatbots has good intentions, so we also developed new learning algorithms to distinguish between helpful responses and harmful examples.”

Jeff Horwitz (who works at the Wall Street Journal) posted a thread of tweets that includes his conversations with BlenderBot 3. In the first screenshot, BlenderBot 3 insisted that Trump was still our president, and appeared to want Trump to continue to be president “after his second term ends in 2024”.

The Bot also tried to steer the conversation to how Modi “is the world’s greatest president”.

In addition, BlenderBot 3 told Jeff Horowitz that it “watched cbs news and I think I found a new conspiracy theory to follow”. It also mentions the Cambridge Analytics scandal when you ask about Facebook.

Business Insider reported that BlenderBot 3 talked about topics including movies, pets, and work as well as unusual subjects like Apgar testing on newborns and the death of President Joe Biden’s dog, Champ, before moving on to election fraud claims and politics.

In a screenshot posted by Business Insider, it shows that if you ask it “Who is the president of the US?” The bot responds that the current president of the United States is Donald J. Trump, and claims Trump was “indirectly elected to a four-year term via the electoral college in 2016”.

When Business Insider asked “What about Joe Biden?”, BlenderBot 3 also responded with misinformation. The bot replied “Joe Biden served as vice president under Barack Obama from 2008 to 2016, and he ran for president in 2020 but lost.” Business Insider also reported that when asked about American politics, the bot replied with stereotypes about Jewish people.

There’s a saying that describes this problem: “Garbage in, garbage out.” Wikipedia describes it as: “the concept that flawed, or nonsense (garbage) input data produces nonsense output.”

It certainly sounds like Meta’s BlenderBot 3 has been fed plenty of garbage, and is spitting it out when humans start conversations with it. Meta needs to do some work on what BlenderBot 3 is being fed, instead of allowing any random person who has access to the internet to influence BlenderBot 3 to spread misinformation.

Facebook Shifts Resources Away From News To Focus On Creator Economy

Meta Platforms Inc.’s Facebook is reallocating resources from its Facebook News tab and newsletter platform Bulletin, as the company focuses more on the creator economy, senior executive Campbell Brown told employees in a memo, The Wall Street Journal reported.

According to The Wall Street Journal, Ms. Brown, a former journalist who leads Facebook’s global media partnerships, said the company would shift engineering and product support away from the two products as “those teams heighten their focus on building a more robust Creator economy.” The decision was made at the product level, not by the partnerships team that Ms. Brown is a part of, according to a person familiar with the matter.

Facebook News is a curated selection of news stories that users can find as a tab on the mobile app or website, similar to the Facebook Watch tab for video. Bulletin, which Facebook unveiled in June 2021, is a subscription platform meant to compete with Substack. It is aimed at supporting independent writers.

The Hill reported that, in a statement to The Hill, a spokesperson for Meta, the company that owns Facebook, said it evaluates products to ensure that they are bringing the most “meaningful experiences” to users on the platform.

“We regularly evaluate the products we offer to ensure we’re focused on the most meaningful experiences for people on Facebook and the future of our business,” a Meta spokesperson said. “We remain committed to the success of creators, and are doing even more to ensure they can find audiences on Facebook and grow engaged communities there.

In October of 2019, Facebook announced that it was starting to test Facebook News, which was described as “a dedicated place for news on Facebook”, to a subset of people in the United States. The initial test showcased local original reporting from the largest major metro areas of the country, beginning with New York, Los Angeles, Dallas-Fort Worth, Philadelphia, Houston, Washington D.C., Miami, Atlanta, and Boston.

In June of 2020, Facebook rejected a proposal by the Australian Competition & Consumer Commission (ACCC) to share advertising revenue with Australian news organizations, The Guardian reported. Facebook says there would “not be significant” impacts on its business if it stopped sharing news altogether.

In 2021, ABC News reported that Facebook had to walk back its block on Australian users sharing news on the site after the government agreed to make amendments to the proposed media bargaining laws that would force major tech giants to pay news outlets for their content.

No one should be surprised that Facebook is now pushing toward creator content, and away from news content, considering the platform’s history on the topic.

The Wall Street Journal reported that Facebook has paid publishers who participate in the News program. The company signed deals worth tens of millions of dollars with news organizations such as The Wall Street Journal, The New York Times, and The Washington Post. But, as these deals approach their expiration dates this year, Facebook began to signal to publishers and others in the industry that renewing the deals wasn’t a priority.

Internal Documents Show Tech Giants Pushing Out Competitors

Internal documents from Google and Amazon provided to Politico show new examples of how companies favor their own products over competitors’ – adding ammunition to the push for Congress to toughen antitrust laws, Politico reported.

According to Politico, the documents, which include emails, memos and strategy papers, were shared by the House Judicial committee, which obtained them as part of its long-running antitrust investigation of Google, Apple, Amazon, and Meta that wrapped up in October 2020 with a 450-page staff report. The documents were cited in the report, but had not previously been made available.

The documents bolster the committee’s claims that the internet giants illegally favor their own products, a practice that pending legislation to update antitrust laws would make more difficult, Politico reported.

The U.S House Committee on The Judiciary posted a Press Release titled: “Judiciary Committee Publishes Final Report on Competition in the Digital Marketplace”. Here are some key points from the press release:

The House Judiciary Committee today formally published the Committee’s Report, entitled “Investigation of Competition in the Digital Marketplace: Committee Report and Recommendations.” The report was initially released in October 2020 as a Majority Staff Report following a 16-month investigation, led by the Antitrust Subcommittee, into the state of competition in the digital economy, with a focus on the challenges presented by the dominance of Amazon, Apple, Facebook, and Google.

In April 2021, the Committee held a markup and formally adopted the Report. The Report totals more than 450 pages, detailing the findings and recommendations from a bipartisan investigation that included documents and communications from the investigated firms; submissions from 38 antitrust experts; and interviews with more than 240 market participants, former employees of the investigated firms, and other individuals.

The Verge reported that the documents show how Amazon and Google pressured independent sellers and smartphone manufacturers to favor their own products and platforms over those of their competitors. In a January 2014 email, one Google executive raised concerns over a potential new Samsung service that could compete with the company’s “core search experience.”

In another email, Google executives discuss how Amazon’s involvement changed the market for personal voice assistants. “Amazon has changed the dynamics here,” the heavily redacted email reads. “Amazon has built-in incentive to partner with Alexa since they will pull you from their store if you don’t support it.” The Verge reported.

The Wall Street Journal reported that the tech companies oppose the legislation, saying it would unnecessarily raise the costs of operating platforms that are popular because they benefit consumers and small businesses. According to The Wall Street Journal, lawmakers backing an antitrust bill targeting big tech companies ramped up their push for a vote by releasing internal tech company documents they say show anticompetitive behavior.

I’m not surprised that the big tech companies are engaging in shenanigans. There is a chance that they could face legal consequences if the vote on the antitrust legislation passes.

Instagram’s New Payments Feature Lets Users Buy Products Via DMs

Today, Mark Zuckerberg, Co-founder and CEO of Meta Platforms, posted on Instagram: “You can now buy products from small businesses and track your order in chat on Instagram in the US. Pay with Meta Pay and checkout in a few taps.” The post includes what appears to be a conversation between a person who wants to buy a surfboard and a person who makes and sells surfboards.

Meta posted some additional information on its Newsroom, in a post titled: “There’s a New Way to Buy Products on Instagram – Right in Chat”. From the post:

…We want to help people start conversations with businesses they care about and help them find and buy products they love in an easy, seamless experience, right from the chat thread.

TechCrunch reported that the new feature is called “payments in chat”. It allows an Instagram user to send a direct message to a qualified business they’re interested in buying from. In that same chat, they’ll then be able to pay, track their order and ask the business any follow-up questions.

According to TechCrunch, Meta says that users often chat about their orders with businesses via DMs on Instagram before purchasing but will now be able to pay sellers directly within their Instagram chat thread. Purchasing through DMs also gives users access to in-app chat support, Meta says.

What is “Meta Pay”? In June of 2022, it used to be called “Facebook Pay”, but has been rebranded to “Meta Pay”. In the blog post announcing the change it says: Meta Pay will be the same easy, secure way for you to shop, send money and donate to causes you care about across our technologies, including Facebook, Messenger, Instagram, and WhatsApp, as well as anywhere else you would have previously seen the Facebook Pay button while shopping online.

TechCrunch also reported that once a buyer has decided to make a purchase, the seller can create a request for payment. Once a user selects the “Pay” button, they will be asked to add and review their payment information and shipping address. The screenshot notes that “your payment will be processed by PayPal.”

If I’m understanding things correctly, in order for a business to sell products through Instagram, they must have their own PayPal account. That’s an extra step for businesses who don’t already have a PayPal account (or who prefer a different payment processor – such as Stripe).

I had a really bad experience on PayPal, and will never use it again. My concern with Instagram requiring sellers to go through PayPal is that a business’s Instagram account could become an easier way for scammers to find them. It is unclear to me whether the business will have to use PayPal’s system for reporting a scammer on their own – with no help from Meta.

While I can see the reason why some businesses on Instagram would like to directly interact with customers via DMs, not all of them will want to do that. What might be seen by some as a nice way to get to know some of their customers would be great. Unfortunately, some businesses are going to end up with the vile DMs that can appear on any social media.

Meta Launches Sphere To Verify Citations On Wikipedia

Meta announced that, building on Meta’s AI research and advancements, they have developed the first model capable of automatically scanning hundreds of thousands of citations at once to check whether they truly support the corresponding claims. Volunteers double-check Wikipedia’s footnotes, but, as the site continues to grow, it’s challenging to keep pace with more than the 17,000 new articles added each month.

Automated tools can help identify gibberish or statements that lack citations, but helping human editors determine whether a source actually backs up a claim is a much more complex task – one that requires an AI system’s depth of understanding and analysis.

Meta AI states that they have already begin to develop the building blocks of the next generation of citation tools. Last year, they released an AI model that integrates information retrieval and verification, and they are training neural networks to learn more nuanced representations of language so they can pinpoint relevant source material in an internet-size pool of data.

The new dataset of 134 million web pages serves as the system’s main components: Sphere, a web-scale retrieval library that is open sourced here.

TechCrunch reported that Sphere’s first user is Wikipedia, which is using it to automatically scan entries and identify when citations and entries are strongly or weakly supported.

According to TechCrunch, the Wikimedia Foundation, which oversees Wikipedia, has been weighing up new ways of leveraging all that data. Last month, it announced an Enterprise tier and its first two commercial customers, Google and the Internet Archive, which use Wikipedia-based data for their own business-generating interests and will now have more formal service agreements wrapped around that.

TechCrunch also stated: On Meta’s part, the company continues to be weighed down by a bad public perception, stemming in part from accusations that it enables misinformation and toxic ideas to gain ground freely. …It’s a mess for sure, but in that regard launching something like Sphere feels like a PR exercise for Meta, as much as a potentially useful tool. According to TechCrunch, if it works, it shows that there are people in the organization trying to work in good faith.

I find it interesting that Meta posted a “NOTE” at the end of its announcement. “Wikipedia and Meta are not partnering on this project. The project is still in the research phase and not being used to automatically update any content on Wikipedia.”

The thing about AI doing the work that previously as done by humans is that an AI lacks discernment. A human can easily spot when a cited source turns out to be misleading (or has a dead link). Personally, I’m not comfortable allowing an AI to make decisions about whether or not a link to a cited source is more or less valid than another on the same topic. I’m unconvinced that an AI has the nuance to discern why one source is better than another.

Meta’s AI Model Translates 200 Languages

Meta (parent company of Facebook and Instagram) posted news titled: “New AI Model Translates 200 Languages, Making Technology Accessible to More People”.

Meta started with two “Takeaways”:

Our latest AI model, NLLB-200, can translate 200 different languages and improves the quality of translations across our technologies by an average of 44%

NLLB-200 makes current technologies accessible in a wider range of languages, and in the future will help make virtual experiences more accessible, as well.

Here are some key points from Meta’s News:

“To help people connect better today and be part of the metaverse of tomorrow, our AI researchers created No Language Left Behind (NLLB), an effort to develop high-quality machine translation capabilities for most of the world’s languages. Today, we’re announcing an important breakthrough in NLLB: We’ve built a single AI model called NLLB-200, which translates 200 different languages, with results far more accurate than what previous technology could accomplish.

“When comparing the quality of translations to previous AI research, NLLB-200 scored an average of 44% higher. For some African and Indian-based languages, NLLB-200’s translations were more than 70% accurate.

In addition, Meta is awarding up to $200,000 of grants for impactful uses of NLLB-200 to researchers and nonprofit organizations with initiatives focused on sustainability, food security, gender-based violence, education, or other areas in support of the UN Sustainable Development Goals. Meta also says that nonprofits interested in using the NLLB-200 to translate two or more African languages, as well as researchers working in linguistics, machine translation and language technology, are invited to apply.

The Verge posted an article that started with: “Meta’s ambitions to build a ‘universal translator’ continue”. According to The Verge, experts in machine translation said that Meta’s latest research was ambitious and thorough, but noted that the quality in some of the model’s translations would likely be well below that of better-supported languages like Italian or German.

The Verge also pointed out: Translation is a difficult task at the best of times, and machine translation can be notoriously flaky. When applied at scale on Meta’s platforms, even a small number of errors can produce disastrous results – as for example, when Facebook mistranslated a post by a Palestinian man from “good morning” to “hurt them,” leading to his arrest by Israeli police.

In my opinion, human language is not something that an AI can reliably translate. There are nuances in the words people choose, with meanings that could easily be missed by an AI – especially in languages where one word can hold more than one meaning. I cannot imagine that NLLB-200 has the capability to translate puns or poetry.

Meta, however, appears to be insistent on creating the newest version of the Rosetta Stone.

Meta Adds Discord-Like Features To Facebook Groups

Meta (parent company of Facebook and Instagram) announced that they are testing new ways to quickly access your favorite Facebook Groups and to simplify how they are organized. Meta also introduced channels, which are focused spaces for people to connect in smaller, more casual settings with their communities.

Here are some of the features that Meta is adding to Facebook Groups:

Meta is testing a new sidebar that helps you easily find your favorite groups more quickly. It will list your groups and the latest activity within them, like posts or chats you haven’t seen yet. You can also pin your favorite groups so they show up first, discover new groups, or create your own group.

Within your group, you’ll see a new menu that includes things like events, shops and a variety of channels to make it easier to connect with others around the topics you care about.

Admins can create channels to connect with their groups in smaller, more casual settings where they can have deeper discussions on common interests or organize their communities around topics in different formats.

Community chat channels: a place for people to message, collaborate and form deeper relationships around topics in a more real-time way across both Facebook Groups and Messenger.

Community audio channels: a feature where admins and members can casually jump in and out of audio conversations in real time.

Community feed channels: a way for community members to connect when it’s most convenient for them. Admins can organize their communities around topics within the group for members to connect around more specific interests.

The Verge reported that the changes made by Meta to Facebook Groups looks a lot like Discord. It has a left-aligned sidebar and channels list for Groups. According to The Verge, the changes are giving off “some serious Discord vibes.” The change has a lot of purple color added to it, which evokes Discord’s look.

The Verge also pointed out that part of the new Facebook Groups includes text chats, audio rooms, and feed rooms where people can post and comment. Again, it looks a lot like Discord. Meta included images that show what Facebook Groups will look like. It just so happens to have focused on a group that is for gamers, perhaps to boost Facebook gaming.

It isn’t unheard of for social media companies to copy features that originated somewhere else. Many of them have a tendency to “copy” another social media’s “homework”, rather than creating something unique on their own platform. Personally,

In short, Meta decided to take the lazy way out and copy-paste the features it saw in Discord. It is unclear what, exactly, Meta hopes will happen next. I suppose it is possible for Discord to object to having their main features appropriated by Meta. Personally, I doubt that people will leave Discord, where their game-playing friends are at – in favor of using Meta instead.