Tech Brew Ride Home - Mon. 11/28 – The China Protests

Episode Date: November 28, 2022

All of the protests erupting in China have, not only a definite tech angle, but might have been triggered by those lockdowns at that Foxconn factory. BlockFi officially files for bankruptcy. Binance attempts to prove its reserves. How to message yourself on WhatsApp. And how Minecraft might have led to a major new breakthrough in AI. Sponsors: LetsTalkInTouch.com Storyblok.com/ridehome Links: Twitter grapples with Chinese spam obscuring news of protests (Washington Post) US Bans Huawei, ZTE Telecom Equipment on Data-Security Risk (Bloomberg) Crypto Lender BlockFi Filing for Bankruptcy and Conducting Major Layoffs as FTX Contagion Claims Another: Source (Decrypt) Binance releases proof-of-reserves system, starting with bitcoin (The Block) WhatsApp rolls out a feature that makes it easier to message yourself (WhatsApp) A bot that watched 70,000 hours of Minecraft could unlock AI’s next big thing (MIT Technology Review) Special Series Launch: The Promises And Perils Of A Decade Of AI Funding (Crunchbase News) Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to the Tech Meme Right Home for Monday, November 28th, 2020. I'm Brian McCullough today. All of the protests erupting in China have not only a definitive tech angle, but might

also have been triggered by those lockdowns at that Foxcon factory. BlockFi officially files for bankruptcy, finance attempts to prove its reserves, how to message yourself on WhatsApp, and how Minecraft might have led to a major new breakthrough and AI. Here's what you missed today in the world of tech. You might remember, I left you last week with news of riots. at that Foxcon factory in China. I'm not exactly saying those riots led to this, but you might have seen over the weekend, protests of various sorts have erupted all across China. Protesters are holding up blank pieces of paper as a protest against censorship, leading platforms and sensors

to quickly remove those images, but also they're fighting a rearguard action all across social media. Researchers say China seems to be flooding Twitter with porn tweets. mentioning places where people are protesting, overwhelming Twitter's reduced moderation team. Quoting the Washington Post. Numerous Chinese language accounts, some dormant for months or years, came to life early Sunday and started spamming the service with links to escort services and other adult offerings alongside city names. The result, for hours, anyone searching for posts from those cities and using the Chinese

names for the locations would see pages and pages of useless tweets instead of information about the daring protests as they escalated to include calls for Communist Party leaders to resign. It is not the first time that suspected government-connected accounts have used the technique, according to a recently departed Twitter employee, but in the past it was used to discredit a single account or a small group by naming them in the escort ads. This is a known problem that our team was dealing with manually, aside from automations we put in place, said the former employee who spoke on the condition of anonymity to avoid retribution for disclosing internal processes. In mass layoffs and resignations, Twitter's overall staff

has been slashed from about 7,500 to roughly 2,000. Surviving employees have estimated. Some groups, including those dealing with human rights issues, safety concerns, and deceptive foreign influence operations, have been reduced to a handful of people or no staff at all. Sunday's campaign was, quote, another exhibit where there are now even larger holes to fill, the ex-employee said, all the China influence operations and analysts at Twitter all resigned, end quote. Meanwhile, back to Foxcon. Sources are saying the turmoil at that huge Foxcon hub is likely to lower iPhone 14 Pro production by close to 6 million units by the end of this year, potentially worsening if lockdowns continue, or I guess if protests continue. But there are bigger things to think about than iPhone

production numbers, you know? Related, I guess, the FCC. has imposed a ban on equipment from Huawei and ZTE, camera providers Hickvision and Da Hua and radio maker Hightera, citing a risk to national security. Quoting Bloomberg. In the four to zero vote, the FCC concluded the products pose a risk to data security. Past efforts to curb Chinese access include export controls to cut off key, sophisticated equipment, and software, and recently U. U.S. officials have weighed restrictions on TikTok over fears Chinese authorities. could access U.S. user data via the video sharing app. This is a culmination action, said Klan Kitchen, a senior fellow at the Washington-based American Enterprise Institute, a public policy think tank.

Things that began under Trump are now being carried out. The Biden administration is continuing to turn the screws on these companies because the threat isn't changing, end quote. In an email response, Hick Vision said its video security products, quote, present no security threat to the United States, and there is no technical or legal justification for the Federal Communications Commission's decision, end quote. The company said, the ruling will, quote, make it more harmful and more expensive for U.S. small businesses, local authorities, school districts, and individual consumers to protect themselves, their homes, businesses, and property, end quote. Huawei declined to comment, well, Da Hua, Hightera, and ZTE, didn't respond to emails sent outside normal business hours in China. Chinese foreign ministry spokesman Zhao Li Jian urged the U.S. to stop what

he described as an attempt to politicize and weaponize economic trade and science technology issues. What the U.S. has done is another example of abusing the concept of national security, using state power to hobble Chinese companies, Zhao told a regular news briefing Monday, end quote. There had been rumbling about it for weeks now, but it finally happened. Today, BlockFi officially filed for Chapter 11 bankruptcy protection in New Jersey and says it plans to lay off a large portion of its staff, also saying that it has $256.9 million left in terms of cash on hand. Quoting to Crypt. BlockFi, which let users earn yield for depositing idle cryptocurrencies on the platform,

first halted withdrawals on November 11th, the same day FTCS filed for bankruptcy. We, like the rest of the world, found out about this situation through Twitter, BlockFi wrote in a letter at the time, We are shocked and dismayed by the news regarding FTX and Alameda, end quote. Nearly a week later, a source at the company told, crypt that it was considering filing for bankruptcy given its hefty exposure to FTX, end quote. Related? We've heard of proof of work and proof of stake, but Binance has released its proof of reserves system, starting with Bitcoin, claiming it has 582,485 Bitcoin, and in terms of user

balance, 575,742 Bitcoin, a reserve ratio of 101 percent, quoting the block. The exchange also provided a link for Binance users to verify their own Bitcoin on the exchange. Binance said it will add further tokens and blockchains in the next few weeks. It also wants to involve third-party auditors to check the proof of reserve system and to implement zero-knowledge-proof technology to provide cryptographic proof of its claims while protecting user privacy. This comes just weeks after rival exchange FTX collapsed, after seemingly swapping user funds for other more illiquid tokens,

eventually leading to a liquidity crisis. Binance's goal is to show that it holds its users' assets in the same tokens that they have deposited, end quote. WhatsApp has begun rolling out a message yourself feature, letting users send notes, reminders, and shopping lists to themselves. Quoting TechCrunch. Sending messages to your own account can be a way to keep a piece of information easily accessible right next to your other WhatsApp conversations. Called message yourself, the feature lets users send notes, reminders, and shopping lists to themselves on WhatsApp. On Monday, the meta-owned instant messaging app announced the rollout of the new messaging feature that will reach all its Android and iPhone users in the coming weeks. It was initially tested with some beta testers.

WhatsApp beta tracker W-A-Beta Info reported in late October. The company has confirmed to TechCrunch that the feature has begun rolling out globally. Users will see their contact at the top of the contacts list on WhatsApp when they create a new chat. Tapping that contact will take them to the chat screen they can use to send messages to themselves. Although the native feature to message yourself is new on WhatsApp, some users have already been using a workaround for some time. You could already send a message to yourself using the app's click-to-chat feature. Nonetheless, the new offering removes the additional steps that users needed to self-chat using the W-A-M-E URL. Users can also pin their self-chat messages to the top

of the conversation list if they don't want to search them in their widely polluted chats list. WhatsApp rival's signal has a feature named Note to Self that addresses the same use case. It lets you create messages for personal use. However, unlike WhatsApp's newly launched feature that is accessible from the top of the contact lists on the app, Signal doesn't suggest your own profile at the top of the recipient list. Users need to search and select the contact entry, note to self, to use the feature. Similarly, community platform Slack has a dedicated space titled Jot Something Down to let users send notes to themselves, end quote. More news in the AI space, MIT Technology Review, takes a look at OpenAI's Minecraft bot,

which has learned to complete complex tasks inside Minecraft after being trained on 70,000 hours of video of people playing Minecraft. But the news here is that this all led to potentially a new breakthrough in AI training. Quote, OpenAI has built the best Minecraft playing bot yet by making it watch 70,000 hours of video of people playing the popular computer game. It showcases a powerful new technique that could be used to train machines to carry out a wide range of tasks by binging on sites like YouTube, a vast and untapped source of training data. The Minecraft AI learned to perform complicated sequences of keyboard and mouse clicks to complete tasks in the game, such as chopping down trees and crafting tools.

It's the first bot that can craft so-called diamond tools, a task that typically takes good human players, 20 minutes of high-speed clicking are around 24,000 actions. The result is a breakthrough for a technique known as imitation learning in which neural networks are trained to perform tasks by watching humans do them. Imitation learning can be used to train AI to control robot arms, drive cars, or navigate web pages. There is a vast amount of video online showing people doing different tasks. By tapping into this resource, the researchers hope to do for imitation learning, what GPT3 did for large language models. In the last few years, we've seen the rise of this GPT3 paradigm where we see amazing capabilities come from big models trained on enormous swaths

of the internet, says Bowen Baker at OpenAI, one of the team behind the new Minecraft bot. A large part of that is because we're modeling what humans do when they go online, end quote. The problem with existing approaches to imitation learning is that video demonstrations need to be labeled at each step. Doing this action, this happen, doing that action, makes that happen, and so on. Annotating by hand in this way is a lot of work, and so such data sets tend to be small. Baker and his colleagues wanted to find a way to turn the millions of videos that are available online into a new data set. The team's approach, called video pre-training, or VPT, gets around the bottleneck in imitation learning by training other neural networks to label

videos automatically. The researchers first hired crowd workers to play Minecraft and recorded their keyboard and mouse clicks alongside the video from their screens. This gave them 2,000 hours of annotated Minecraft play, which they used to train a model to match actions to on-screen outcomes. Clicking a mouse button in a certain situation makes the character swing its axe, for example. The next step was to use this model to generate action labels for 70,000 hours of unlabeled video taken from the internet and then train the Minecraft bot on this larger data set. Video is a training resource with a lot of potential, says Peter Stone, executive director of Sony AI America, who has previously worked on imitation learning. Imitation learning is an alternative to reinforcement learning in which a neural network learns to

perform a task from scratch via trial and error. This is the technique behind many of the biggest AI breakthroughs in the last few years. It has been used to train models that can beat humans at games, control a fusion reactor, and discover a faster way to do fundamental math. The problem is that reinforcement learning works best for tasks that have a clear goal. where random actions can lead to accidental success. Reinforcement learning algorithms reward those accidental successes to make them more likely to happen again. But Minecraft is a game with no clear goal. Players are free to do what they like, wondering a computer-generated world,

mining different materials, and combining them to make different objects. Minecraft's open-endedness makes it a good environment for training AI, end quote. And finally, this is related to that. According to a crunch-based news analysis, AI startup funding has been around 9 to 10% of global VC investment in recent years, led by robotics, autonomous vehicles, and enterprise software. Quote, leading the way in AI startup funding are robotics, autonomous vehicles, and enterprise software, but other sectors have quickly followed.

Biotechnology, computer vision, language processing, fintech, semiconductors, security, logistics, learning agtech, and more. Back in 2013, funding to companies using AI was $3 billion, with $3 billion, with fewer than 1,000 deals. In 2021, AI funding peaked at $69 billion across more than $4,000 deals. Venture investment in the technology has reached $38 billion so far this year. The most highly valued AI companies on the Crunch Base Unicorn Board are data and AI company Databricks valued at $38 billion, driverless auto company Cruise at $30 billion, research and development company OpenAI at $20 billion, and AI writing assistance service

grammerly at $13 billion. In this current investing climate, valuations are likely down from these peaks in 2021. The number of AI companies that raised their last funding at seed since 2020 have raised more than $1 million in equity adds up to more than $600 every year. Of those companies, Series A makes up more than 400 companies per annum, and at Series B, that count is 200 companies. In most recent times, cybersecurity startups saw a record $2.1 billion in venture investment last year, according to crunch-based data. The pandemic also sparked new interest in AI-enabled medical diagnostics with funding for such startups jumping from $348 million in 2020 to more than $1 billion in 2021. In fact, AI has crept into the most surprising of sectors, improving everything from

how we get our sleep, farm our fish, and design our wigs. Who knows what's next? End quote. Absolutely wild day over at the World Cup this morning. If Cameroon v. Serbia wasn't enough for you with that 3-3 comeback. I hope you caught Ghana's win over South Korea, because the end of that one was just chaos. Football, man. Talk to you tomorrow.