Tech Brew Ride Home - Mon. 12/05 – Stack Overflow Bans ChatGPT

Episode Date: December 5, 2022

Maybe put the breaks on an early release of that Apple AR/VR headset. And maybe pump the breaks on the AI excitement as Stack Overflow temporarily bans users from sharing responses generated by ChatGPT. A security issue at Rackspace. Are the Winkivii in trouble? And why has Sam Bankman Fried not been arrested yet? Sponsors: LetsTalkInTouch.com Hover.com/ride for 10% off Links: Kuo: Apple Headset Shipments Potentially Delayed Until Second Half of 2023 (MacRumors) Foxconn sees COVID-hit China plant back at full output in late Dec-early Jan -source (Reuters) Rackspace rocked by ‘security incident’ that has taken out hosted Exchange services (The Register) AI-generated answers temporarily banned on coding Q&A site Stack Overflow (The Verge) Crypto broker Genesis owes Winklevoss exchange’s customers $900mn (Financial Times) Why Hasn’t Sam Bankman-Fried Been Arrested Yet? (Intelligencer) 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, December 5th, 2020. I'm Brian McCullough today. Maybe put the brakes on expectations for an early release of that Apple AR VR VR headset. And maybe pump the breaks on the AI excitement as well as Stack Overflow temporarily bans users from sharing responses generated by

chat GPT. A security issue at Rackspace are the WinkleVye in trouble. And why has Sam Bankman Freed not been arrested yet? Here's what you missed today in the world of tech. Well, maybe got to put a bit of a break on the excitement we let off with on Friday. Our buddy Ming Chi Kuo has a research note out saying that mass shipments of Apple's ARVR headset may be delayed until the second half of 2023. That would be a delay from the previously estimated Q2 of 2020 launch. Quo says the delay would be due to software-related issues, quoting Mac rumors. Quo said mass shipments of components for the headset are still likely to begin in the first half of 2023, but he believes that mass shipments of fully assembled headsets may not begin until the second half of the year.

This delay could result in the headset launching in the second half of 2023, even if Apple announces it earlier in the year. WWDC 2020 in June would be an opportune time for Apple to announce the headset, along with an SDK for developers to create ARVR apps for the device. Quo previously claimed that Apple planned to unveil the headset at a January media event and begin mass shipments in the second quarter of 2023. Quo is unsure whether the delay would result in the January announcement being pushed back, but he did note that an overly lengthy period between a media event and mass shipments can hurt sales.

Apple's headset is expected to be an expensive niche product, with Quo forecasting that the company will ship fewer than 500,000 headsets in 2023, end quote. Following up on another recent Apple headline, a source is telling Reuters that Foxcon expects its Zhangzhou plant to resume full production in late December or early January, but plans may change November revenue at Foxcon apparently dropped 11.4 percent, quote, at present, the overall epidemic situation has been brought under control with November being the most affected period. The company said in a statement, adding it has started to recruit new employees and was gradually, quote, restoring production capacity to normal, end quote.

Foxconn said November revenue for its smart consumer electronics business, which includes smartphones, declined year-on-year, partly due to a portion of shipments being impacted by production disruptions in Zhangzhou. It did not elaborate. Following the November unrest that saw workers clash with security personnel, Foxcon could have seen more than 30 percent of the Zhangzhou's cites November production affected. Reuters reported last month citing a source familiar with the matter. Foxcon hasn't disclosed details of the impact of the disruption on its production plans or finances. The capacity is now being gradually resumed, with new staff hiring underway, said the person with direct knowledge of the matter. The person declined to be named as the information

was private. Quote, if the recruitment goes smoothly, it could take around three to four weeks to resume full production, the person said, pointing to a period around late December to early January. A second Foxcon source familiar with the matter said the company is hoping to resume full production as soon as possible but was not able to give a timeline. Quote, the situation has stabilized, the person said, referring to the protests and the government's easing of COVID restrictions. the local government is actively helping with the resumption, end quote. Rackspace has had to confirm a, quote, security incident impacting its hosted Microsoft Exchange services, leading to an extended outage. Customers have been complaining all weekend of poor

customer support surrounding this issue, quoting the register. Some of Rackspace's hosted Microsoft Exchange services have been taken down by what the company has described as a security incident. The company's most recent incident report at the time of writing, timestamped 157 Eastern time, December 3rd, offers the following information. On Friday, December 2, 2022, we became aware of an issue impacting our hosted exchange environment. We proactively powered down and disconnected the hosted exchange environment while we triaged to understand the extent and severity of the impact. Upon further analysis, we have determined that this is a security incident, end quote. The incident is further described as isolated to a portion of our hosted exchange platform,

end quote. Rackspace has no idea when it will be able to restore its service to those impacted by the security incident. Rackspace has offered impacted customers free access to Microsoft Exchange Plan 1 licenses on Microsoft 365 for the duration of the incident and shared instructions on how to get that up and running. The instructions suggest the work to get it running will take 30 minutes to an hour. The register has received accounts of the incident from customers trying to access Rackspace support. They're not pretty. Quote,

The way they have handled this has been horrible. No support, no email, no nothing. Who knows when we will have answers, one customer wrote. I called the support line, held, and listened to lousy music for three hours and 14 minutes and 19 seconds, and finally had to terminate the call, wrote another. We've also seen reports that the process of migrating to Office 365 is not straightforward, as it may require information sent to email addresses that are inaccessible because of Rackspaces out. end quote. And maybe pump the breaks on all the AI excitement, or maybe not, because maybe this

headline can be read as bullish as it could show disruption already happening because of these AI advances. Stack Overflow has temporarily banned users from sharing responses generated by chat GPT, pending a ruling in the future. Most answers provided by chat GPT are incorrect, at least according to the powers that be at Stack Overflow, quoting the verge. The site's mods said that the ban was temporary and that a final ruling would be made sometime in the future after consultation with its community. But as the mods explained, chat GPT simply makes it too easy for users to generate responses and flood the site with answers that seem correct at first, but are often wrong on close

examination. The primary problem is that while the answers which chat GPT produces have a high rate of being incorrect, they typically look like they might be good, and the answers are very easy to produce. Wrote the mods, as such, we need the volume of these posts to reduce, so for now, the use of chat GPT to create posts here on Stack Overflow is not permitted. If a user is believed to have used chat GPT after this temporary policy is posted, sanctions will be imposed to prevent users from continuing to post such content, even if the posts would otherwise be acceptable, end quote. Chat GPT is an experimental chat bot created by OpenAI and based on its autocomplete text generator, GPT3.5.

A web demo for the bot was released last week and has since been enthusiastically embraced by users around the web. The bot's interface encourages people to ask questions and in return offers impressive and fluid results across a range of queries, from generating poems, songs, and TV scripts to answering trivia questions and writing and debugging lines of code. But while many users have been impressed by chat GPT's capability, abilities, others have noted its persistent tendency to generate plausible but false responses. Ask the bot to write a biography of a public figure, for example, and it may well insert incorrect biographical data with complete confidence. Ask it to explain how to program software for a specific function, and it can similarly produce believable, but ultimately incorrect code.

This is one of several well-known failings of AI text generation models, otherwise known as large language models or LLMs. These systems are trained by analyzing patterns in huge reams of text, scrape from the web. They look for statistical regularities in this data and use these to predict what words should come next in any given sentence. This means, though, that they lack hard-coded rules for how certain systems in the world operate, leading to their propensity to generate fluent bullshit. Given the huge scale of these systems, it's impossible to say with certainty what percentage of their output is false. But in Stack Overflow's case, the company has judged for now that the risk of misleading users is just too high. Stack Overflow's decision is particularly notable,

as experts in the AI community are currently debating the potential threat posed by large language models. Jan Lecun, chief AI scientist at Facebook Parent Meta, has argued, for example, that while LLMs can certainly generate bad output like misinformation, they don't make the actual sharing of this text any easier, which is what causes harm. But others say the ability of these systems to generate text cheaply at scale necessarily increases the risk that it is shared later. To date, there has been little evidence of the harmful effects of LLMs in the real world, but these recent events at Stack Overflow support the argument that the scale of these systems does indeed create new challenges. The sites mods say as much in announcing the ban on chat GPT, noting that the volume

of these AI-generated answers, thousands, and the fact that the answers often require a detailed read by someone with at least some subject matter expertise in order to determine that the answer is actually bad, has effectively swamped our volunteer-based quality curation infrastructure, The worry is that this pattern could be repeated on other platforms, with a flood of AI content drowning out the voices of real users with plausible but incorrect data. Exactly how this could play out in different domains around the web, though, would depend on the exact nature of the platform and its moderation capabilities. Whether or not these problems can be mitigated in the future using tools like improved

spam filters remains to be seen, end quote. Hopefully not more signs of contagion in the crypto space, but sources are telling the financial Times that Gemini, the crypto exchange owned by the Winklevoss Twins, is trying to recover $900 million owed to it by Genesis Global and Genesis's parent company, Digital Currency Group. Quote, Genesis is the main partner in Gemini's Earn Program, where retail investors lend out their coins in exchange for a fixed stream of returns. Gemini halted withdrawals from the scheme last month after Genesis said, quote, unprecedented market turmoil meant it did not have sufficient liquidity to make good on all its redemption requests. Gemini has now formed a creditors committee to recoup the funds from Genesis

and its parent DCG, the people said. Gemini and Genesis declined to comment. Genesis has been scrambling to raise funding and has hired investment banking boutique Moles and company to help it explore all possible options, according to the people familiar with the situation. The creditor committee is in negotiations with both Genesis and DCG, the parent group of Genesis, which is run by billionaire Barry Silbert, the people said. DCG was founded in 2015 and is one of the biggest investors in the crypto industry. It was valued at $10 billion last year by investors including Singapore's sovereign wealth fund, GIC, Google's venture arm, Capital G, and SoftBank, and its subsidiaries include Genesis and Investment Manager Grayscale.

DCG itself owes money to its subsidiary genesis. These inner company loans have complicated the picture for creditors. DCG has $2 billion worth of outstanding debt, $1.7 billion of which which is owed to its own subsidiary Genesis through two loans. Over the summer, Genesis lost $1.1 billion on a loan to collapsed hedge fund Three Arrow's capital. DCG took on Genesis's liabilities in the process, subsequently owing $1.1 billion to Genesis. Silbert told investors last week that DCG had separately borrowed $575 million from Genesis, quote, on an arm's length basis, end quote, to fund undisclosed investments and share buybacks from non-employee shareholders, end quote.

And finally today, if you're in the crypto space, you don't need me to tell you that Sam Bankman-Fried has continued his all-out publicity blitz, agreeing to be interviewed by All in Sundry, including lowly podcasts and YouTube channels. So people all weekend have been asking the question, why is he able to do this? Why has he not been arrested yet? Well, in Intelligencer, Ancuch-Cardori makes the plausible case that, unlike, say, Bernie Madoff, SBF has not yet confessed to any criminal conduct. arguing instead that FTC's collapse was purely sloppiness on his part. That likely means a U.S. prosecution could take some time to come together. Quote, let's start with the actual circumstances of Bernie Madoff's arrest. The scheme was not detected by the government or the media, infamously, but was in fact revealed by Madoff himself who confessed the whole thing to his sons once he realized that he could not keep up the fraud. According to the criminal complaint filed on the day of his arrest. Madoff told his sons, quote, that he was finished, that he had absolutely nothing, that it's all just one big lie, and that it was basically a giant Ponzi scheme, end quote.

His sons called the FBI, and two days later, two agents showed up at his home and asked whether, quote, there's an innocent explanation for what Madoff had told his sons. Madoff literally replied, quote, there is no innocent explanation, end quote. At the risk of stating the obvious, the reason Madoff was arrested so quickly is because he confessed to every element of criminal fraud, including both the underlying scheme and his criminal intent. This meant that the FBI had both that confession and highly potent admissible evidence of guilt in the form of testimony from his adult children who had no apparent acts to grind. We have not seen anything like a real admission of criminal conduct from SBF yet,

and so far as I can tell, he has held firm on a central point for his defense, that the epic still unspooling fiasco at FDX was the result of sloppiness and inadvertent missteps by the company's leadership rather than an intentional effort to mislead FTX customers or investors. This means that the Justice Department will need to try to get to the bottom of it by conducting an investigation that could take a long time, which is not to say that it will take SBF's claim at face value. In fact, I am sure it will not. Since his comments on the most prominent issue at the moment, FTX's transfer of customer funds to SBF's hedge fund, Alameda Research, and the subsequent loss of more than $1 billion, have been transparently evasive and substantively meaningless.

We have seen this before, down to eerily familiar media minutia. In 2016, just days after the Wall Street Journal published its first expose about Theranos, Elizabeth Holmes took the stage for an interview at a conference organized by the paper in which she insisted that the whole thing was a huge misunderstanding. I certainly understand why people would bristle at all those sources of potential delay, but this sort of investigative slog is much more common in a white-collar case than a quick arrest. For instance, it took almost a year for the U.S. Attorney's Office in Manhattan, which is reportedly handling the FTX investigation, to charge the founder of the electric truck startup Nicola with fraud, even though the short-selling firm Hindenberg Research seemingly did all

of the work when it published a lengthy report that precipitated the investigation in the first place. And it took nearly two years for the department to charge homes, even after the journal's John Carreou publicly exposed the central problems at Theranos. Prosecutors at the Justice Department cannot and do not simply indict people based on what they read in the press, no matter how damning it may seem. Even if they desperately want to charge someone, prosecutors have to conduct their own investigation to develop robust, admissible evidence of criminal misconduct, which requires gathering and reviewing documents and data, speaking with witnesses, and perhaps above all, time, end quote. Over the weekend, we showed the kids, Indiana Jones, and the last

crusade for the first time, and I was annoying everybody with my Sean Connery impressions, but As I kept on with it, my wife said, I was drifting into a Daniel Plainview impression. From there will be blood, so I guess this weekend I discovered I can do a Daniel Plainview? You tell me, I run a family business. This is my son and partner, H.W. If I tell you I'm an oil man, you'll believe me. God damn hell of a show. I don't know. Close. Anyway, talk to you tomorrow.