1. Home
  2. Meta.

Meta.

Meta launches suite of generative AI music tools rivaling Google’s MusicLM

Meta’s latest suite of generative AI tools allows users to input text to create musical and audio compositions, rivaling a similar tool released this year by Google.

Meta, the parent company of Facebook and Instagram, launched a suite of generative AI models on Aug. 2, which it calls AudioCraft, for the purpose of music creation from various inputs, according to a blog post

Included in the suite of generative AI tools are MusicGen and AudioGen, which operate off of text-based inputs to create new audio, along with another called EnCodec that “allows for higher quality music generation with fewer artifacts.”

In the announcement, Meta mentioned that its MusicGen model was trained with music that it owns or “specifically licensed.”

This comes as there have already been major controversy surrounding training AI with copyrighted work across many artistic fields, including a lawsuit against Meta for copyright infringement during AI training.

At the moment, Meta has made MusicGen and AudioGen available in several sizes to the “research community” and developers. It said as it develops more advanced controls it envisions the models to become useful to both amateurs and professionals in the music industry.

“With even more controls, we think MusicGen can turn into a new type of instrument — just like synthesizers when they first appeared.”

In a recent interview between Cointelegraph and the CEO of the Recording Academy, Harvey Mason Jr., he also likened the emergence of AI-generated music to the early days of synthesizers coming onto the music scene.

Related: Spotify reportedly deletes thousands of AI-generated songs

Meta’s release of its generative AI music tools comes shortly after Google launched its set of similar tools that turns text into music called MusicLM. 

In May, the company announced that it was accepting “early testers” of the products via its AI Test Kitchen platform.

Meta has been actively releasing new AI tools, alongside many other tech giants including Google and Microsoft, in a race to develop and deploy the most powerful models.

On Aug. 1 Meta announced the launch of new AI chatbots with personalities, which users on its platforms will be able to use as search helpers and as a “fun product to play with.”

Magazine: BitCulture: Fine art on Solana, AI music, podcast + book reviews

Bitcoin price slips to $93K as liquidations soar and long-term BTC holders take profit

Alibaba releases two open-sourced AI models to rival Meta’s Llama 2

The Chinese tech and e-commerce giant Alibaba revealed two new open-sourced AI models, Qwen-7B and Qwen-7B-Chat, to rival Meta’s recently-launched Llama 2.

Alibaba Group, the Chinese tech and e-commerce giant, announced the release of two open-sourced artificial intelligence (AI) models from its cloud computing department on Aug. 3, according to a press release. 

Its two large language models (LLMs) are dubbed Qwen-7B and Qwen-7B-Chat, each with 7 billion parameters. Alibaba said these two models are small-size versions of the Tongyi Qiawen, which the company released in April.

The new models aim to help introduce AI to the operations of small and medium-sized businesses.

The company said Qwen-7B and Qwen-7B-Chat have various capabilities that would be appealing to enterprises, such as being able to “code, model weights, and documentation will be freely accessible to academics, researchers and commercial institutions worldwide.”

Alibaba’s latest LLMs are also the first released from a Chinese tech company to be open-sourced. However, it said businesses with over 100 million monthly active users will need a license.

On Aug. 1, the company also announced an update in the form of a vector engine to its AnalyticDB data warehousing service, allowing its corporate clients to quickly create custom generative AI applications.

Related: AI chip developer gets $100 million from Samsung and Hyundai

This development comes after Meta released its open-sourced LLM — Llama 2 — with Microsoft on July 16.

Meta says its Llama 2 was trained using 40% more public data and can process twice as much context as its predecessor. It is also open-sourced, with the biggest version of Llama 2 featuring 70 billion parameters.

Similar to Alibaba’s latest model, it requires a license from companies with over 700 million monthly users.

On July 26, Alibaba announced what it called “the first training and deployment solution for the entire Llama2 series in China” after it deployed a Llama 2 solution for businesses to develop AI-powered software and tools.

Magazine: Make 500% from ChatGPT stock tips? Bard leans left, $100M AI memecoin: AI Eye

Bitcoin price slips to $93K as liquidations soar and long-term BTC holders take profit

Meta to launch AI chatbots with personalities to retain users: Report

As the Big Tech giant Meta focuses on retaining its users it plans to launch a number of AI-powered chatbots with different personalities and functions.

Meta, the parent company of Facebook and Instagram, plans to release artificial intelligence (AI) chatbots with human-like personalities as it focuses on user retention efforts, according to a report from the Financial Times. 

On Aug. 1 the FT reported that people close to the matter said prototypes of the chatbots have been underway with the final products being able to hold discussions with users at a human level.

The range of chatbots will be able to display different personalities and are expected to be released as early as next month, according to the report.

Sources close to the matter say that Meta staffers have dubbed the chatbots as “personas” and that these bots take the form of different characters. Another person with knowledge of the plans said the company has already explored one bot that speaks like the former United States president Abraham Lincoln and another for travel advice that speaks like a surfer.

According to the sources from the FT, the purpose of the chatbots will be both to offer recommendations and new search functionality, along with being a “fun product for people to play with.”

Cointelegraph has contacted Meta for further comment on the matter and has received no response at the time of writing. 

Related: Meta and Microsoft launch open-source AI model Llama 2

The FT source said that the company may automate checks on the chatbots’ outputs to ensure accuracy and avoid rule-breaking speech. 

This development comes as Meta has allocated major efforts toward user retention. During its 2023 second-quarter earnings call on July 26, Meta CEO Mark Zuckerberg spoke about the company’s latest product and Twitter rival, Threads.

He said that he is “seeing more people coming back daily than I’d expected.” It was in this call Zuckerberg said Meta was primarily focused on Threads user retention. The earnings call also revealed another $3.7 billion invested into metaverse development.

The introduction of accessible chatbots also means an opportunity to collect large amounts of user data. OpenAI, the maker of the viral AI chatbot ChatGPT, has been hit with a class-action lawsuit over alleged data theft via its own bots. 

Magazine: Experts want to give AI human ‘souls’ so they don’t kill us all

Bitcoin price slips to $93K as liquidations soar and long-term BTC holders take profit

Meta Q2 earnings: Reality Labs losses top $7.7B year to date

Meanwhile, Meta's metaverse-building business has racked up around $21 billion in losses since the start of 2022.

Meta's metaverse-related losses topped $3.74 billion over the second quarter with the Big Tech player spending $7.7 billion on its virtual reality business so far in 2023.

Its second-quarter 2023 results released on July 26 saw Meta report an 11% revenue gain compared to the same quarter last year, totaling $31.9 billion.

Its metaverse-focused Reality Labs revenue topped $276 million, its lowest in two years and a nearly 40% drop compared to Q2 2022.

Meta's segment results in millions since Q2 2021 with added highlights on Reality Labs' Q2 2023 revenue and operating losses. Source: Meta

On an earnings call, Meta's financial chief Susan Li said Reality Labs' revenue drop was due to lower sales of its Quest 2 virtual reality (VR) headset. The department's expenses were up 23% to $4.0 billion partly caused by a growth in staffing costs.

Reality Labs' operating losses are set to increase through 2023, Meta said. It cited VR-related product development efforts and further investments in its metaverse as the reason for the losses extending.

On the call, Meta chief Mark Zuckerberg said the firm is focusing on artificial intelligence "in the near term and the metaverse over the longer term."

He reiterated Meta is "fully committed" to its metaverse alongside its AI investments and said the two areas are "overlapping and complementary."

He added its AI model Llama is being used to build a number of products that will help users "create worlds and the avatars and objects that inhabit them as well" and said he would share more later in the year.

Related: ‘Already explored’ — Apple Vision Pro fails to impress Mark Zuckerberg

Meta's stock price jumped on the earnings and is up over 7% in after-hours trading to around $320 according to Google Finance data. Meta shares have gained nearly 140% year-to-date but are still off from their September 2021 all-time high of over $378.

Meta's stock price spiked to over $320 in after-hours trading on July 26. Source: Google Finance

Zuckerberg mentioned its July 6 launched platform Threads was “seeing more people coming back daily than I’d expected” and said Meta was focused firstly on Threads user retention, then growth and would later focus on monetizing the platform.

The comments come the same day as a July 26 report from data analytics firm Similarweb that claimed Threads users have declined 60% from launch.

Threads peaked at 49 million daily active users on July 7 but fell to 12.6 million daily active users by July 23 with users spending less than five minutes a day on the app over the past week.

Web3 Gamer: Apple to fix gaming? SEC hates Metaverse, Logan Paul trolled on Steam

Bitcoin price slips to $93K as liquidations soar and long-term BTC holders take profit

Meta Platforms fined $14 million for Onavo privacy issues: Report

The decision in an Australian court was driven by increasing concerns about illicit activities in the crypto market and the desire to safeguard investors.

An Australian court has reportedly ordered Meta Platforms, the owner of Facebook, to pay fines amounting to 20 million Australian dollars ($14 million) for collecting user data through a smartphone application, Onavo. 

According to a Reuters report, the Federal Court of Australia has directed Meta, along with its subsidiaries Facebook Israel and the discontinued app, to reimburse $270,356 (A$400,000) in legal costs to the Australian Competition and Consumer Commission (ACCC). The ACCC initiated the civil lawsuit against Meta, alleging that Onavo was promoted as a privacy protection tool, but failed to openly reveal its data collection methods.

Facebook used Onavo to collect users' location, time and frequency using other smartphone apps and websites they visited for its own advertising purposes, Judge Wendy Abraham said in a written judgment, according to the report.

Related: Alibaba to support Meta’s AI model Llama: Report

Meta reportedly stated that the ACCC had recognized their lack of intent to mislead customers and they emphasized their efforts in developing tools over the past few years to provide users with increased transparency and control over their data usage.

The imposed fine marks the conclusion of one aspect of Meta's legal challenges in Australia concerning its management of user data, Reuters said. This legal matter emerged amid a scandal involving Meta's association with data analytics firm Cambridge Analytica during the 2016 United States election.

However, Meta's legal woes are not over yet, as it is reportedly also facing a civil court action by Australia's Office of the Information Commissioner regarding its dealings with Cambridge Analytica specifically in Australia.

Cointelegraph reached out to Meta for more information, but had not received a response bthe time of publication.

Magazine: Is the Metaverse really turning out like ‘Snow Crash’?

Bitcoin price slips to $93K as liquidations soar and long-term BTC holders take profit

Alibaba to support Meta’s AI model Llama – report

Alibaba’s cloud computing service taps Meta’s artificial intelligence model Llama to power new software development.

E-commerce group Alibaba will become the first Chinese company to make use of Meta’s open-source artificial intelligence (AI) model Llama to power zero-cost development of programs.

According to an initial report from Reuters which cited an official statement from Alibaba Cloud through its WeChat account, the cloud computing arm of the Chinese conglomerate has deployed a Llama 2-based solution to allow businesses to develop software and tools using AI:

“Today, Alibaba Cloud has launched the first training and deployment solution for the entire Llama2 series in China, welcoming all developers to create customised large models on Alibaba Cloud.”

Meta’s Llama2 model was released in July 2023 as a free-to-use service rivaling the likes of early movers including OpenAI’s ChatGPT and Google Bard. Meta intends to allow Llama2 to be free to use for companies with less than 700 million monthly active users.

The announcement from Meta highlighted that Microsoft remained its preferred partner for developing its generative AI tool while Llama 2 would be freely available for research and commercial use.

Related: OpenAI launches official ChatGPT app for Android

Meta also noted that it was adopting an open approach to provide increased access to foundational AI technology to benefit businesses around the world. This includes supporting companies around the world that are building products on Llama 2, cloud providers including the model in its offerings as well as research efforts exploring “safe and responsible deployment of large generative models”.

Alibaba Cloud is set to become one of a handful of prominent cloud computing services including Amazon Web Service (AWS) that will tap into Llama 2’s large language model.

The development comes after the United States moved to curtail the sale of select AI processing hardware chipsets in June 2023 as the country looks to retain a competitive advantage in the rapidly developing area of AI tools.

The Reuters report also suggests that the integration of Meta’s new AI model could present an avenue for the country to reaffirm ties in China. Meta’s Facebook has been banned in China since 2009 alongside Twitter, YouTube and other Western-based social media and content platforms.

Cointelegraph has contacted Alibaba Cloud and Meta to ascertain the finer details of the Llama 2 integration, but hasn't yet received a response.

Collect this article as an NFT to preserve this moment in history and show your support for independent journalism in the crypto space.

Magazine: Experts want to give AI human ‘souls’ so they don’t kill us all

Bitcoin price slips to $93K as liquidations soar and long-term BTC holders take profit

Trademark attorney predicts legal fights ahead with Twitter’s rebrand to X

According to Josh Gerben, the social media platform could be embroiled in lawsuits over its 'X' branding for years, costing upwards of millions of dollars.

Josh Gerben, a trademark attorney with Gerben Intellectual Property, has suggested the company formerly known as Twitter could spend upwards of $100 million in dealing with litigation as a result of the X rebranding.

In a July 25 X thread, Gerben predicted that lawsuits against X related to trademark infringement could appear in United States courtrooms “in the next few weeks” with legal problems on the international stage likely to be a “very big issue” for years. U.S.-based companies including Microsoft and Meta already own similar ‘X’ trademarks for different products and services, and many others may have grounds for a lawsuit against X.

“To cast aside an asset this valuable [the blue bird] in favor of a new trademark is unprecedented in history,” said Gerben. “If Elon had stuck with the Twitter brand, his company would likely not spend a dime to justify its Twitter trademark on just about anything he wanted to do with the brand.”

Former Twitter CEO Elon Musk announced the rebranding would be taking place on July 23, with the company’s iconic blue bird being replaced by a black and white ‘X’ logo. X Corp is the parent company of Twitter, established by Musk in March. The platform’s search bar still bears its former namesake at the time of publication, as does the blue ‘tweet’ button and retweet counter.

Many on X have criticized the rebranding as not well thought out. Musk himself was unable to provide a definitive answer as to the new name of a retweet in a July 24 thread. He has also hinted that users should expect to conduct their “entire financial world” on the X platform with plans to offer services affecting “half of the global financial system”.

The move is the latest controversy by one of the biggest social media platforms in the world. Musk took over Twitter in October 2022 following a $44-billion agreement, later firing members of the company’s executive team and phasing out the platform’s legacy verification system — blue checkmarks — in favor of various paid options.

Related: Reading the Elon Musk ‘tea leaves’ — Is Dogecoin coming for Twitter?

Though Musk briefly changed the Twitter logo to that of the Dogecoin (DOGE) shiba inu in April, many on the platform speculated at the time that the decision was only temporary. For the rebranding to X, the company’s logo on its brick and mortar headquarters in San Francisco may soon change from a bird to a letter:

Under Musk, U.S. advertising revenue for Twitter reportedly dropped roughly 59% from April to May compared to that over the same period in 2022. Former NBCUniversal exec Linda Yaccarino took over as CEO in May, while Musk continues to actively post Xs on the social media site.

Magazine: Musk hints at suing Microsoft, US Rep. wants Gensler fired, and more: Hodler’s Digest, April 16-22

Bitcoin price slips to $93K as liquidations soar and long-term BTC holders take profit

Ark Invest sells more Coinbase shares, eyes Meta platforms, Robinhood

Cathie Wood, the Bitcoin advocate, is actively investing in Meta Platforms (META) and Robinhood Markets (HOOD) shares.

As the stock price of Coinbase (COIN) reaches a 52-week high, Ark Invest, the investment management firm led by Bitcoin (BTC) bull Cathie Wood, has recently divested more of its Coinbase shares.

Following Ark Invest's significant acquisition of Coinbase shares during the bear market, the investment firm has recently sold off a portion of its Coinbase holdings for the second time this week. In the meantime, Cathie Wood, the prominent Bitcoin advocate, is actively investing in Meta Platforms (META) and Robinhood Markets (HOOD) shares.

Trade information observed by Cointelegraph reveals that on July 14, Cathie Wood's Ark Invest funds divested a total of 478,356 Coinbase (COIN) shares valued at $53 million, coinciding with the stock's yearly high of $114.43.

Screenshot of  ARK Invest trading notification. Source: ARK Trading desk.

In the transactions, Cathie Wood's flagship fund ARK Innovation ETF (ARKK) divested 263,247 COIN shares, ARK Next Generation Internet ETF (ARKW) sold 93,227 COIN shares, and ARK Fintech Innovation ETF (ARKF) offloaded 35,666 COIN shares.

In June, Cathie Wood initiated the purchase of Meta Platforms shares after the company announced the launch of "Threads," a social media app similar to Twitter. ARK Innovation ETF (ARKK) acquired 69,793 META shares, while ARK Fintech Innovation ETF (ARKF) purchased 111,843 shares of Robinhood. Additionally, ARK Next Generation Internet ETF (ARKW) increased its holdings with 12,559 META shares and 169,116 Robinhood shares.

As per CoinMarketCap, the closing price of COIN stock on Friday stood at $105.31, experiencing a 1.58% decline as investors secured their profits. During the week, COIN price surged by 33%, reaching a 24-hour high of $114.43, and displaying an impressive year-to-date increase of 213%. In addition to the overall upswing in crypto-related stocks, Coinbase witnessed a notable rise following the summary judgment in the US SEC v. Ripple lawsuit.

Related: Allowing Coinbase to go public was not a ‘blessing’ from regulators — SEC

Cathie Wood’s ARK Innovation ETF (ARKK) sold 135,152 Coinbase (COIN) shares worth $12 million on July 11 and ARK Fintech Innovation ETF (ARKF) sold 160,887 Coinbase (COIN) shares in March.

Magazine: How smart people invest in dumb memecoins: 3-point plan for success

Bitcoin price slips to $93K as liquidations soar and long-term BTC holders take profit

Report: Meta to release commercial AI tools to rival Google, OpenAI

Sources close to Meta reportedly have said the company plans to make a commercial version of its AI model to be more widely available and customizable.

Meta plans to release a commercial version of its artificial intelligence (AI) model aiming to reach wider usage according to a Financial Times report.

The report contained details from sources close to Meta who said that although the company already released its own language model called LLaMa for researchers and academics earlier this year, the new version will be more widely available and can be customized by companies.

This comes as Meta tries to position itself to be competitive with Microsoft-backed OpenAI, the creator of ChatGPT, and Google who are current leaders in the space.

FT’s source, reported to have knowledge of high-level strategy at Meta, said:

“The goal is to diminish the current dominance of OpenAI.”

With the commercial version of LLaMa, start-ups and businesses will be able to build custom software and applications on top of Meta’s underlying AI technology.

At the moment all of Meta’s models are free and open-source though two of FT’s sources said the company has been exploring a paid version for enterprise customers. However, it would not be a part of the upcoming release.

The release of the commercial version is expected “imminently,” says the FT sources close to the situation.

Related: Elon Musk kickstarts new AI company to ‘understand the universe’

Moreover, Meta has made its LLM models open-source which means the details of the system's operations are publicly available. This is not the case with its competition like OpenAI, which keeps its code private to third parties.

FT’s source close to Meta said:

“Meta realized they were behind on the current AI hype cycle, and this gives them a way to open up the ecosystem and seem like they are doing the right thing, being charitable and giving back to the community.”

Cointelegraph reached out to Meta for additional comments on the development and its strategy. 

These dvelopements come as Meta faces a lawsuit from author Sarah Silverman, along with two other authors on behalf of a class of copyright owners across the United States alleging that Meta has committed copyright infringement while training its AI systems. 

OpenAI has been hit with a similar class-action lawsuit th alleges the company has committed data theft in the training of its models. 

Magazine: How smart people invest in dumb memecoins: 3-point plan for success

Bitcoin price slips to $93K as liquidations soar and long-term BTC holders take profit

Sarah Silverman sues Meta and OpenAI for copyright violations

Author Sarah Silverman and two others opened a lawsuit against OpenAI and Meta for using copyrighted work without permission to train their AI systems.

The American comedian and author Sarah Silverman, along with two other authors Richard Kadrey and Christopher Golden, have filed lawsuits against Meta Platforms’ LLaMa and OpenAI’s ChatGPT over copyright infringement. 

Meta and OpenAI are alleged to have used the plaintiffs’ content for training their respective artificial intelligence (AI) systems without obtaining any prior permission.

According to the court documents against Meta, many of the plaintiffs’ books under copyright appear in the dataset that “Meta has admitted to using to train LLaMA.”

Similarly, in the case against OpenAI, the lawsuit alleges that when ChatGPT generates summaries of the plaintiffs’ work it is an indication of the training via copyrighted content.

“The summaries get some details wrong. This is expected since a large language model mixes together expressive material derived from many sources. Still, the rest of the summaries are accurate…”

In order to obtain this data the suits claim that the companies retrieved the copyrighted data from what are known as “shadow libraries,” such as Bibliotik, Library Genesis, Z-Library, and others.

Related: Japanese AI experts raise concern over bots trained on copyrighted material

These shadow libraries are websites that use torrent systems to make books “available in bulk," says the lawsuit. Such sites are illegal and are unlike open-source data that comes from databases such as Gutenberg, which collects books that have copyrights that have run out.

“These shadow libraries have long been of interest to the AI-training community because of the large quantity of copyrighted material they host.”

Along with complaints about copyright infringement of their own personal work, the authors filed the complaint on behalf of a class of copyright owners across the United States whose works were also allegedly infringed. 

Cointelegraph reached out to OpenAI and Meta for comment on the case, though neither responded prior to publication.

In May writers across the U.S. a part of the Writers Guild of America, took to the streets in an authorized strike -the first one in 15 years- which highlighted many issues faced in the industry including the usage of AI.

Magazine: Super Mario: Crypto Thief, Sega blockchain game, AI games rights fight — Web3 Gamer

Bitcoin price slips to $93K as liquidations soar and long-term BTC holders take profit