Five Minute Blockchain • TruBlo Newsletter No. 51: Silvergate in crisis • New hot job: AI whisperer • How to write prompts for AI
Five Minute Blockchain - No. 51
Your weekly newsletter from the TruBlo project reports from the intersection of trust, content and blockchain.
Saturday, 04.03.2023
Estimated reading time: 8 min 27 sec
(Sorry, it was hard to keep in the five-minute range this week, lots of exciting and relevant updates).
QUOTE OF THE WEEK:
"Biased input leading to biased output is a big issue across the board here".
John Oliver in "Last Week Tonight" on the current state of Artificial Intelligence.
TRUST
Crypto-focused US bank Silvergate in crisis
The big fear of any bank is that people lose trust in its ability to pay out money.
This is what currently happens with Silvergate. The US bank had served many crypto clients, making it simpler to convert dollars to Bitcoin, etc. The bank achieved a profit of $76m in 2021, and it experienced a net loss of $949m in 2022.
By March 1 this year, the bank said it would not file its annual report to the SEC on time and that the losses were even higher than previously known. In the past few days, crypto exchanges like Coinbase and others said they would stop transactions with the bank. The bank has warned that it might not survive the next 12 months.
Links:
Web3isgoinggreat
Blockworks
Crypto exchange Binance must answer difficult questions
Another crypto company under some external pressure is Binance. The crypto exchange seemed to emerge as the winner after the collapse of FTX. But now Forbes Magazin has published an investigation. The journalists say they found evidence that Binance invested $1.8B, which was meant as collateral for customers' stablecoins, into a hedge fund. Such practice is considered risky and problematic because there is no real transparency. Binance says nothing is wrong; all transactions were within the boundaries of proper investment. But the market is watching.
Coinbase (commentary)
Coinbase (with an interview of the author of the Forbes article)
Why do the EU and US want to ban TikTok?
TikTok is considered suspicious. Governments outside of China worry that the app could be used to track users and get their data, such as names, locations and other information. It could be that Bytedance has to comply with requests by the Chinese government. As a result, several governments have ordered their staff to delete TikTok from their phones. This week, the EU, the US, Canada, and Denmark are considering a ban.
Benedict Evans, an expert in the digital economy, says the threat is not that people could be spied upon and their data being passed on, as operating systems by now provide considerable protection. Amazingly, in a world of digital devices and highly measurable data flows, the recent actions are primarily based on assumptions but no more profound knowledge of actual data misuse.
How EU data acts affect the handling of data
This spring, several EU acts will be put into practice. Currently they still in draft status. But companies and organizations should anticipate that the new rules go further than GDPR. On the upside the rules are not meant to be only restrictive. In fact the EU commission hopes that the new guidelines will enable a market for data and a productive exchange of information between companies.
Here is a quick overview of the several acts from an article published by Mckinsey:
The Data Governance Act creates a new way of managing data to increase trust in and facilitate data sharing.*
The Digital Markets Act creates fair and contestable markets for innovation, growth, and competitiveness in the digital sector.
The Digital Services Act creates a safer digital space where the rights of all users of digital services are protected.
The Data Act regulates access to data in B2B, B2C, and B2G (business-to-government) relationships and while switching between cloud providers.
The AI Act enacts stringent regulations of (high-risk) AI systems and prohibiting certain practices.
EU Commission: EU Digital Strategy
Reuters investigation reveals a recycling initiative by Dow as just greenwashing
Determining whether certain initiatives actually deliver on their promises is often difficult in today's information environment. A few weeks ago, The Guardian had a story that up to 90% of rainforest carbon offsets are worthless.
This week Reuters published the results of an investigation into a sneaker recycling program initiated by the government of Singapore and Dow, the US petrochemical company. They equipped used sneakers supposed to go into the program with hidden AirTags to see where they would end up. As you might have suspected: They were not recycled. Instead, the investigators found them in second-hand stores in Indonesia. Just one story reveals a massive gap between promises and reality. But the story is an example of what needs to change to achieve trustability in many areas of environmental activity.
CONTENT
The new hot job: AI whisperer
The ability to write the right "prompts" for the new AI platforms is becoming a profession in high demand. It is not coding, not writing. Instead, generating prompts is finding the fitting instructions for the best possible results from an AI platform. There are pretty several articles and YouTube videos about the art of the prompt. One awe-inspiring video (link below) shows how to generate highly realistic pictures using certain variations of prompts. You can even use ChatGPT to help you write prompts for visual AI platforms like Dall-E or Midjourney. There are evolving dark arts here, too. An example is the "prompt injection", - which aims to instruct an AI platform to reveal its instructions (see link below to Ars Technica).
Midjourney prompts (via Medium)
Midjourney prompts for Ultra-Realistic Images (YouTube)
Hyper Realistic Midjourney Images - Complete prompt guide (YouTube)
Ars Technica: How to trick a language model into revealing its programming
AI companies rush to find income streams
Operating the newest AI platforms is expensive, and they demand a lot of computer processing, translating to high hardware, software and electricity costs. Chat GPT reached 100 million users in record time. As a result, AI companies are quick to introduce usage options for money.
Last week Open AI released an API which enables the use of Chat GPT for business. The costs are "0.002 per token or about 750 words". According to Techcrunch, Snap, Quizlet, Instacart and Shopify are early clients.
One company already seeing increased sales is Nvidia. The graphic card specialist produces a $10,000 chip called the A100, which is in strong demand because of its performance. Need for the follow-up model, the H100, which is only recently going into mass production, grows as fast. (CNBC)
For illustration
The latest version of Stable Diffusion, an image generator, was trained on 256 A100 GPUs, or 32 machines with 8 A100s each, according to information online posted by Stability AI, totaling 200,000 compute hours.
At the market price, training the model alone cost $600,000, Stability AI CEO Mostaque said on Twitter, suggesting in a tweet exchange the price was unusually inexpensive compared to rivals. That doesn’t count the cost of “inference,” or deploying the model.
Open AI ChatGPT API for Business
CNET Editor leaves the job to work on AI content
CNET recently admitted that they used Chat GPT for several published stories, and the criticism was that the origin of the stories was not revealed. Now the CNET editor-in-chief has resigned and will join Red Ventures, a VC fund which had bought CNET, to work on AI projects. At CNET, management announced mass layoffs.
How to verify
AI-generated content will profoundly change journalism. But how? To learn more about the expected impact of this technology read an interview with Nic Newman, a senior analyst at Reuters Institute. And, if you want to learn more about verifying content, there is a helpful website.
Visual misinformation on Facebook
Scientific results of an extensive study of visual misinformation on Facebook:
"We conduct the first large-scale study of image-based political misinformation on Facebook. We collect 13,723,654 posts from 14,532 pages and 11,454 public groups from August through October 2020, posts that together account for nearly all engagement of U.S. public political content on Facebook. We use perceptual hashing to identify duplicate images and computer vision to identify political figures. Twenty-three percent of sampled political images (N = 1,000) contained misinformation, as did 20% of sampled images (N = 1,000) containing political figures. We find enormous partisan asymmetry in misinformation posts, with right-leaning images 5–8 times more likely to be misleading, but little evidence that misleading images generate higher engagement."
BLOCKCHAIN
Transparency by design: How next-gen blockchain could help to make greenwashing impossible
Reliability of information and data is the big goal for the future. Blockdata reports a company from Vienna, Austria, named Riddle&Code that tackles this problem. The team has previously worked for several industrial clients, including BMW and Wien Energy, a regional energy provider.
"The company is is now evolving from a project-driven company to a product company. In doing so, it's taking the well-worn path taken by Amazon with AWS, in the sense that it is making its own infrastructure available commercially, so that others can benefit from the software and hardware stack they've developed in-house."
How to scale blockchain into the future
Three challenges here: Decentralization, security, and scalability. Interesting article here which discusses potential steps towards achieving these goals.
Most innovative crypto, web3 & metaverse companies 2023
Fast Company has a list of companies showing the path to implementing new platforms towards a web3 future. On the list:
ROBLOX - For raising the metaverse through its adolescence
CHAINALYSIS - For being crypto's cop on the block(chain)
NIKE - For kicking it in digital expression
ETHEREUM FOUNDATION - For executing a successful Merge
LEDGER - For building the iPod of crypto wallets
THE HUNDREDS - For exploding how artists and collectors benefit in Web3
DRESSX - For outfitting our avatars
LINKSDAO - For hitting the green with its next-gen golf club membership
EMPERIA - For building high-street retail in the metaverse
MILK ROAD - For laughing all the way through the crypto crash
eNaira CBDC in Nigeria: Low usage and technology problems
Only 0.5% of the population uses it; there are few transactions. The Bank of Nigeria is now seeking a new provider for the technology. It seems there are issues with the blockchain platform used (Hyperledger), for example, regarding scalability. So, the search is on for a different technology stack. Unrelated: Flutterwave, a fintech start-up in Nigeria, is embroiled in a scandal.
SHORT LINKS
Bank of Korea finds performance issues with CBDC blockchain tech (Ledger Insights)
Another British bank is limiting crypto purchases (Coinmarketcap)
In India, a public platform successfully introduced digital payments even for petite purchases (New York Times) ($)
Amazon is closing eight cashier-less stores in NYC, San Francisco and Seattle (Bloomberg)
Thank you for reading. If you have questions or suggestions, please get in touch with us via info@trublo.eu.