Archive

The World

Biden signs Executive Order on a new EU-U.S. Data Privacy Framework

…The Executive Order introduces new binding safeguards with the aim to address all the points raised by the Court of Justice of the EU (Schrems II decision), (i) limiting access to EU data by US intelligence services to what is necessary and proportionate to protect national security and (ii) establishing a Data Protection Review Court. The European Commission will now prepare a draft adequacy decision, as well as launch its adoption procedure.

The framework thereby establishes a new two-layer redress mechanism, with independent and binding authority:  

First layer: EU individuals can lodge a complaint with ‘Civil Liberties Protection Officer’ of the US intelligence community.

Second layer: individuals will have the possibility to appeal the decision of the Civil Liberties Protection Officer before the newly created Data Protection Review Court. The Court will be composed of members chosen from outside the US Government, appointed on the basis of specific qualifications.

Previously individuals could turn to an Ombudsperson, which was part of the US State Department and did not have similar investigatory or binding decision-making powers. The EU will have 6 months to consider, but privacy advocates are expected to challenge the framework as being inadequate. One such person, Max Schrems said that he, through his privacy advocacy group Noyb will challenge the proposal.

Artificial Intelligence

WHO releases Florence, its AI powered chat bot designed to provide reliable information on health, including Covid

…it is usable in English, with Arabic, French, Spanish, Chinese, Hindi, and Russian to follow.

There are other types of health chat bots. This article mentions the following:

• Ada Health, Healthily and Sensely – combines symptom checkers with information
• Woebot, OneRemission – provide psychological and behavioural support to patients with specific health concerns
• Babylon Health and Gyant – appointment management and connecting patients remotely with clinicians.

Also, Microsoft’s AI can predict who is likely to miss a doctor’s appointment. It can be used to remind patients and schedule appointments so that doctors’ practices can account for no-shows

Former Google engineers release character.ai which enables you to talk to figures, such as Trump, Xi, Musk…

…there is clear warning to users that it is all made up. It’s a bit like the end of a film, reminding everyone that the story is fiction, so the explanation goes.

New AI uses spotted this week

• Now Google’s Imagen can do Text to Video AI – type in a description and a video based on the description will be generated. Check out Imagen Video.  Meta released a similar tool recently.
• Alphabet’s DeepMind has a new AI system, AlphaTensor which can do some serious matrix multiplications speedily. These are complex and key for engineering and physics simulations.

BigTech/ Data / Platforms

Business

Microsoft – Meta partnership: Microsoft’s Teams, Office, Windows and Xbox Cloud Gaming (including the “Netflix of games” Game Pass) to be delivered via the Oculus

…this is a clever partnership. Both parties announced very similar visions of the Metaverse at similar times, about this time last year. Yet they have decided to collaborate, which will no doubt accelerate the concept. Meta’s internal notes revealed this week that it was struggling with take up even within Meta’s employees. The article notes Microsoft gets to trade on what it does best (purveyor of software, especially productivity tools) and Meta gets to access enterprise clients. But thought leader Ben Thompson says Meta may be in danger of becoming what IBM was to Microsoft all those years ago – IBM invested all the R&D in the hardware (PCs) only for Microsoft to benefit from it.

Content Moderation

Now you can influence your feed on Facebook

…you can get the Facebook algorithm to show more or show less on a post to ensure you have a better chance of getting what you want. At the same time, you show more of yourself to Facebook, which enable them to ad-target you better, and incidentally perhaps Facebook can better defend themselves should any content moderation issues crop. Users though will have the benefit of seeing more of what they want, and get better targeted ads. Some users will see it as a win win. Last week I reported on a case which looks at YouTube’s liability for using AI which resulted in a recommendation of a post uploaded by a terrorist organisation. If the recommendation were controlled by the users, the platform could have more of a defence? It might also be less likely to be attacked for content moderating inappropriately, if users had a hand in formulating its own algorithm.

Twitter expands its crowd-sourcing fact checking measure Birdwatch

…misinformation, perpetrated deliberately or not, is rife on Twitter. To combat that, Twitter provided for Birdwatch, a group of people who have proven that they are capable of writing helpful notes. They will then be monitored to see how helpful their contributions are.  Birdwatch algorithm looks at consensus across those with differing viewpoints. This will increase the chances of identifying misleading posts for removal. Users can rate Birdwatch notes as Helpful or Not Helpful. If a Birdwatch member’s scores dip, then they could get their Birdwatch status stripped. Twitter will now be making the Birdwatch fact check tweets visible to all users.

Roblox, Discord, Meta and Snap sued for suicide attempts of girl

…apparently caused by predators, despite various measures which are said to be available on the platforms – such as the scanning of images and GIFs to identify instances of child exploitation. This is a US case, but it comes in the wake of the UK Coroner’s finding that Meta, Pinterest, and other social media platforms are legally to blame for the death of 14-year-old Molly Russell who was exposed to self-harming images and posts as a result of the platforms’ recommendations.  

The present case though concerns the lack of monitoring of messages. On the one hand, users consider that platforms should not have access to messages and on the other hand, platforms are said to be liable for failing to identify illegal and harmful conduct. Striking the balance between privacy and child protection is notoriously difficult and the EU is debating what to do – privacy concerns have been raised against the draft Regulation to Prevent and Combat Child Sexual Abuse which would allow national courts to issue “detection orders,” requiring messaging and hosting companies to actively monitor and files for signs of child sexual abuse and grooming.

One of the complaints was that the platform provided no age verification. This is likely to come soon enough it has been said – not just on porn sites, which was originally targeted in the UK (and then withdrawn owing to potentially fuelling blackmailing activities in case of data breach) but to all sites accessible by children in the UK, EU, US, Australia and beyond, from search engines to social media.

Privacy

Dutch Court rules that forcing employees to turn their webcam on throughout the working day is violation of European Convention on Human Rights

…the decision was made against Chetu, a US company which had hired a Dutch telemarketer. Chetu also had requested the employee to share screen as well. The employee refused and was sacked. The Dutch Court ruled that the business’ requirement conflicted with the respect for the privacy of the workers (Art 8).

Cloud Computing

Google ramps up Cloud offering

…these include:

• Translation Hub: enterprise-scale translation AI Agent for self-serve document translation, available in 135 languages
• Vertex AI Vision: makes powerful computer vision and image recognition AI more accessible to data practitioners. It also reduces the time to create computer vision applications from days to minutes at one-tenth the cost of current offerings.
• Advanced security tools: including Confidential Space will help unlock the value of secure data collaboration. See also the provision of Google Cloud Ready – Sovereign Cloud (which it says complies with EU regs)

And lots more!

Google cloud has also announced new relationships with Coinbase [see below], Prudential plc, Rite Aid, Snap, T-Mobile, Toyota, and Wayfair.

Connectivity

EU Commission thinking about making BigTechs pay for infrastructure upgrades as the world tries to get 5G deployed

…56% of global traffic were derived from just six companies Amazon, Apple, Google,  Meta, Microsoft and Netflix so why not make them pay [and incidentally they are all US companies, not EU] for the upgrade from the copper networks to fiber, to help the struggling telecom companies?

The BigTechs say they are already investing in data centres, subsea cables and helped telecom companies to store content local to deal with traffic spikes – and that besides it’s the telecom companies fault for failing to properly monetise on their infrastructure.

Crypto

Cryptoexchange FTX partners with Visa to enable users to buy items using cryptocurrencies with debit cards

…Visa already has partnerships with Coinbase and Binance, and others. Mastercard has also partnered with Coinbase on NFT. Merchants can now easily deal with users wanting to buy with crypto – Visa and FTX essentially deal with the backend. FTX chief Sam Bankman-Fried said it will likely be used most widely in countries where inflation has spiked, such as Turkey (83%) and Argentina (78%). Yikes.

Google partners with Coinbase to enable customers to pay for its cloud services using cryptocurrencies

…Coinbase will get a commission from the transaction [which is a bit ironic because one of the attractions of cryptocurrencies is the removal of bank commission charges]. Coinbase will move its cloud supplier from AWS to Google Cloud.

BNY Mellon now licensed to carry out crypto custody service

…fund managers can use one custody platform for traditional and crypto currencies. Basically, crypto is getting more mainstream. See also above.

Cybersecurity

Former Uber executive convicted on charges for covering up data breach

…Sullivan (a former lawyer) had be found to have paid $100,000 as reward under the bug bounty program and asked the hackers to sign up to a non-disclosure agreement not to disclose the incident [ he had managed to track down the hackers – and meanwhile they went on to hack other businesses]. The hackers had stolen the data of 50 million customers and 7 million drivers. The conduct took place when the Federal Trade Commission was investigating Uber. New Uber CEO Dara Khosrowshahi found out and fired him. This is the first time an executive has been found guilty for covering up a hack. Sentencing is due which could involve imprisonment, up to 8 years. 

CISOs (chief investigation security officers) are now concerned that it is never acceptable to pay the ransom fee to keep stolen data private.

EVs/AVs

Ford and GM follows Tesla by taking fate into their own hands by securing supply of raw material for batteries

…Tesla leads the trend. Two years ago, it purchased the right to mine Lithium in Nevada, and last month it revealed that it was considering building a refinery in Texas. He will be applying for tax credits made available by the Inflation Reduction Act. This is available for those that are sourcing the minerals from the US or free trade partners and are not having these extracted, processed or recycled by a “foreign entity of concern”. The Australia’s Jervois Global has opened a new Cobalt mine in Idaho which will be helped by the Act. Note that the EU Commission is unhappy about the Act – because it discriminates against EV makers that do not produce in the US, and so is thinking about initiating an action before the WTO.

Batteries for EV requires rare metals, such as Cobalt (Co), Lithium (Li), Manganese (Mn), Nickel (Ni). Prices for these have soared, as has the demand for EV.  But the following make it difficult to contain price rises for the rare metals:

• Geopolitics [especially the Russia/Ukraine war]
• Child labour involved in mining these
• Environmental issues

Other car makers are also implementing measures to secure these materials:

• GM has agreed to pre-pay Livent, a lithium mining group $200million to secure supply
• Ford is paying Liontown Resources to develop a lithium mine.
• Stellantis has invested in Vulcan Energy Resources which has plans to produce lithium in Germany.
• Nikola has purchased battery supplier Romeo Power.

Recall that GM is following Tesla in other ways too – recently it announced that it was designing its own chips.

Rio Tinto plans $500million investment to mine titanium and scandium to reduce China dependency

…these materials are used in aerospace, medical products and fuel cells. 75% of finished products and 61% of scandium are produced in China. Canada has attracted attention of others as well:

• Cathode producers such as Germany’s BASF, Belgium’s Umicore and Korea’s Posco are apparently planning to invest in Canada too.
• Canada has received love calls from German and South Korean governments to secure supplies of Nickel and Cobalt for use in German cars and Korean batteries.

Tesla says it will bin ultrasonic sensors and rely on camera only

…Ultrasonic sensors were primarily used for short-range detections utilised for parking assist, auto summon and collision detection. The former two features will not be available for Teslas that drive using camera only for now.

Note that this is opposite of what Volvo are doing, which is to put laser technology (LiDAR) on all new models. Tesla have never relied on LiDAR, with Musk claiming that anyone relying on LiDAR is doomed.

Metaverse/ VR / AR / MR/ XR

The rise of the digital influencers

…the Chinese BigTechs, ByteDance (TikTok parent), Alibaba and Tencent have poured in cash into the “virtual idol industry” as they have proven to be safer bets compared to human influencers. Virtual idols are much less scandal prone, age-defying, and their IP is easy to control. SoftBank and US fund Sequoia are also investing, and brands such as Bulgari and Pandora are utilising virtual influencers. 

Will it become vogue in the western world too? I think perhaps not, or not as much – the manga culture isn’t as pervasive in the west as it is in the east. But we shall see.

Semiconductors

US implements new export controls restricting sales of semiconductors with US technology

…unless sellers obtain an export licence (which it is understood is difficult to obtain). Targeted Chinese companies cannot now be supplied with hardware or software which contain US technology (this is the so-called “foreign direct product rule” pioneered by the Trump administration – and includes non-US made chips provided US technology is implicated). The aim is to stop China from progressing its AI, supercomputers and other high performance computing capabilities that could be used to develop military applications in particular nuclear and hypersonic weapons. US companies which do business in China are hard hit (Nvidia, AMD, Applied Materials etc). Note there is a carve out – for US and its allies who are making chips in China (eg. Apple – but they are quickly shifting that to other parts of the world, such as India and Vietnam), albeit Korean companies (allies of the US) are concerned that its chipmaking capability in China will be hit. It fears it will not be able to upgrade its lithography plant and its memory plants in China may be affected.

Note also that it might not be that simple, just to effectively ban the export of certain products to China. The assembling of those chips, installing them to devices etc are generally done in China. So who will do that job? Further still, clever corporate structuring involving shell companies may enable Chinese companies to defeat export controls. There are Chinese companies that have a good level of capability and they will just get more funds no doubt from the Chinese government. China boasts the biggest chip subsidy programme. Those Chinese chip makers will also gain more business from Chinese EV companies too, which will in turn allow them to flourish.  Therefore, US measures may not be as effective as it might be hoping.

Headline in Tech news of the week

US Supreme Court to decide the scope of law removing liability for harmful third party content on platform

…Platform companies are shielded by Section 230 of the Communications Decency Act which generally protects websites from liability over content uploaded by users or third parties even when they disseminate illegal, violent conduct or materials, especially if the company has acted diligently removing illegal material.

In this case, the family of the victim of a shooting in Paris by the terrorist organisation ISIS sued YouTube/Google for recommending videos uploaded by ISIS which concerned brainwashing and recruiting new members, which it claimed breached the Antiterrorism Act. The Ninth Circuit said that YouTube was shielded by s.230. The family says s.230 can give platforms protection only insofar as a particular complaint seeks to treat that service as a publisher.

In Twitter v. Taamneh, there is a similar issue – the question there is whether Twitter could be liable because they could have been more proactive about taking down terrorist content.

Had this been a newspaper, then it would likely be found liable. Newspapers have full control of, and so are responsible for the content on its own media. But, is a platform responsible because their algorithms decide what should be on the recommendations page?

Recently, the Texan Federal Court ruled that state law banning content moderation (excludes instances where material uploaded is illegal) by platforms of at least certain size were not unconstitutional whilst a Court in Florida came to a different conclusion. It would seem that a review by Supreme Court would be necessary to clarify the position.

Artificial Intelligence

White House identifies 5 principles in its Non-binding AI Bill of Rights

…The framework applies to (1) automated systems that (2) have the potential to meaningfully impact the American public’s rights, opportunities, or access to critical resources or services.

The 5 principles identified are

• Safe and Effective Systems – Automated systems should be developed with consultation from diverse communities, stakeholders, and domain experts to identify concerns, risks, and potential impacts of the system. Test the product to ensure it is safe and effective and minimise harms.

• Algorithmic Discrimination Protections – proactive and continuous measures should be implemented to protect individuals and communities from algorithmic discrimination.

• Data Privacy – implementation of protection from abusive data practices via built-in protections and ensuring users’ agency over how data about the user is used. Consent should only be used to justify collection of data in cases where it can be appropriately and meaningfully given.

• Notice and Explanation – notifying users that an automated system is being used and explain how and why it contributes to the outcomes.

• Human Alternatives, Consideration, and Fallback – provision of opt out from automated systems in favour of a human alternative, where appropriate

Some states have parallel laws which deal with these concepts.

Meta discloses its text to video AI system Make-a-Video

…Meta said it will use its Responsible AI Framework to evolve its approach. The Framework considers 5 pillars which are:

• Privacy & Security – careful use over face and speech recognition, use of AI to improve privacy                                  
• Fairness & Inclusion – detecting certain forms of statistical bias, AI-driven Portal Smart Camera designed to accurately focuses on people on-camera regardless of apparent skin tone or gender presentation
• Robustness & Safety – working on developing deepfake detection models and multimodal hate speech detection
• Transparency & Control – be more transparent about when and how AI systems are making decisions that impact the people who use Meta’s products, to make those decisions more explainable, and to inform people about the controls they have over how those decisions are made
• Accountability and Governance – give people a way to appeal and seek additional human review of a broad range of content-takedown decisions, which are sometimes first made with the assistance of AI systems

These principles very much overlap with the AI Bill of Rights just published.

New AI uses spotted this week

• Mayo Clinic shows its AI algorithm can pick out patients with atrial fibrillation.
• Microsoft’s AI can predict who is likely to miss a doctor’s appointment. It can be used to remind patients and schedule appointments so that doctors’ practices can account for no-shows.

BigTech/ Data / Platforms

Business

Google to change its search format into a more visual system for the TikTok Generation

…Google search works almost perfectly well, right? But Google must not sit on its laurels for the TikTok generation are more and more carrying out searches on TikTok, rather than Google. According to Google’s own research, almost 40% of young people will search places to lunch in TikTok or Instagram. This is really an incredible shift in habit. Certain Google searches, like that for a travel destination will present various types of results, such as weather, images and videos of sight-seeing spots, how to get there. You can also carry out searches more easily using your camera – take a photograph of your maths homework and Google will solve it for you, for example.

Content Moderation

Advertisers pull plug on Twitter as adverts displayed alongside child exploitation accounts

…YouTube has been there, done that too – when an Argos ad popped up next to a video of  someone emulating kidnapping. But that was 5 years ago. In the case of Twitter, the illegal accounts were not subtle, featuring words such as rape and teens, trading teens/child. The article notes that Twitter has become a hive of adult content (which Twitter does allow), although a very clear line is drawn between adult and child sexual content, latter being illegal.  However Twitter’s policing is reportedly not adequate.

Crypto/NFTs

Argentine low cost carrier Flybondi to issue tickets in NFTs

…NFTs are essentially a tamper proof receipt. It says you own the thing. If that is a ticket, you can prove that you own the real thing. The difference between an NFT ticket and your standard e-ticket is that you can trade the former knowing it be a real ticket. Contractual terms can be embedded to ensure that the original ticket issuer (eg. Flybondi) gets a commission upon transfer. If the ticket is then re-sold, the issuer gets a further cut and so on. Can be used for concert tickets and other events as well.   The Flybondi ticket will allow you to change the name on the ticket etc, of course. There may be a possibility of using stablecoins in the future [and talking of which see below].

Other interesting NFT use cases I have heard about

• Software licence: I have also heard that you could use NFTs for software licences as well. You buy a 10 year licence, but suppose your company becomes insolvent. You can sell off that licence to other businesses.
• Debt: If you are in debt, you could sell tokens for that debt. If the company survives and does well you could profit from that token. Bitfinex (the Crypto exchange) did this by issuing LEO tokens. Thanks to the increase in bitcoin price increase and recovery of lost $4billion after a hack, the price of LEO tokens went up and lenders got their money back.  

Stripe (the leading online digital payments processing service provider) to deal in USDC stablecoin

…these are cryptoassets, but much less volatile (hence the term stablecoin). It means the unbanked can join the economy and those that get paid in a currency which is not their local currency can duck the exchange rate commissions and the payment processing lag. The latter is incredibly important as engineers and developers in countries such as India, other Asian territories and Easter European territories are increasingly engaging in work for western companies (notably the US).

Apple now supporting NFT trading and reportedly insisting 30% commission

…note though Apple levies 15% commission for developers that generate under $1million in revenue. Apple’s position is simple. NFTs are digital assets, so buy NFT using the iOS platform, then 30% commission should be levied (although quite how the cryptocurrency transactions are carried out on iOS I am not sure).  Note that Apple’s move to support NFT trading could signal a tipping point on NFT popularity. This is because the great majority of NFT trades occur on a desk top and not on mobile phones.  

However, there are question marks over whether Apple’s proposal would work. Suppose an NFT on the NFT marketplace OpenSea is being traded. Note that OpenSea trades are often in secondary markets, in which case OpenSea merely brokers the transaction (a bit like Ebay in this regard, they match-make seller and buyer). All OpenSea does is to connect the digital wallets, they will process the transaction, take the 2.5% fee, and the NFT is transferred. Apple is saying that they want the 30% of the entire transaction from OpenSea, not the seller. In this case, OpenSea will lose money.

EV / AVs

Leading LiDAR company Velodyne acquires AI company specialising in providing traffic solutions

…the pair have been working together for some time to provide smart city applications using LiDAR technology. LiDAR is laser sensor technology that is used to enable vehicles to sense the environment around them.  

Gaming

Google terminates its cloud gaming platform Stadia

… Google ploughed in large swathes of cash into cloud based gaming, to build a service marked with low latency and fidelity – meaning users could play great games on low spec devices (because all the complex high graphics rendering is done on the server end). However, the Achilles’ heel, being the internet connection at the users’ end was too much, as those technical advantages were lost. Early on in the project, Google closed down its own gaming studio, leaving it too dependent on others to provide it with the top titles. There simply was not enough content, which meant that Google didn’t have the traction it needed to get Stadia properly off the ground. Developers that had created games especially for Stadia are now left on the lurch. Some commentators have said that Google’s habit of mothballing projects often at an early stage could mean future developers may be wary of developing works for Google.

Stadia technology will be recycled for use in other initiatives such as AR.

WWE (World Wrestling Entertainment) and Take-Two (gaming publisher) have to pay tattoo artist for including wrestler Randy Orton’s tattoos in WWE 2K video-game series

…defences such as fair use and licenced use were rejected. Similar case is pending in Ohio.

Green technology

University of Illinois Chicago discovers a way to convert virtually all carbon dioxide from industrial exhaust into ethylene – critical material for making plastics

…the process uses electrolysis, and the conversion rate is better than previously achieved.

Semiconductors

US is likely to impose tighter restrictions on exporting cutting edge products to China

…especially chip technology. The US aims to prevent China from using its technology for military use, supercomputers and AI applications and to cement US’ position as leading technology developer.

US chipmaker Micron to receive $320m worth of subsidies from Japan

…the aim is to reduce dependence on China and Taiwan amid geopolitical tensions.

US chipmaker Micron to invest $100billion to build a chip factory in New York state

…with the help of US subsidy.

One more thing…

What is art? Tyga x MSCHF collab piece “Wavy Baby” causes beef with Old Skool TM owner Vans

..this is the question before one Second Circuit panel. Tyga [appears to be a singer] and MSCHF [pronounced mischief  – it’s a product studio] dropped [another term for launched these days] the Wavy Baby sneaker which resembles Vans’ sneakers. One party relies on freedom of expression, saying it’s art or parody. The other says TM infringement, and that the First Amendment does not allow a distortion of TM.

The World

FT analysis shows concern over potential China/Taiwan conflict at all time high

…analysis based on annual regulatory filings. Technology sector (especially as they are dependent on semiconductors) most concerned, then Healthcare. US likely to apply more pressure on key suppliers to ban exports to China, it is reported.  

It is probably no coincidence that Apple has said that it will set up a factory in India. India has just announced a chip making incentive scheme, (which Foxconn (who assembles iPhones for Apple) said will be taking up) and it has the second largest number of smartphone users (though I don’t think iPhones dominate there).

Artificial Intelligence

EU Commission proposes new Liability rules for those users harmed by the use of AI

….there are in fact two proposals:

• Revised Product Liability Directive: Modernises the existing rules on the strict liability of manufacturers for defective products (from smart technology to pharmaceuticals). The commission says that it will ensure that victims can get fair compensation when defective products, including digital and refurbished products, cause harm. Eg. compensation for loss arising out of cybersecurity vulnerability.

• AI Liability Directive: Harmonisation of national liability rules for AI – making it easier for victims of AI-related damage to get compensation. The Commission explains: where a relevant fault has been established and a causal link to the AI performance seems reasonably likely, the so called ‘presumption of causality‘ will address the difficulties experienced by victims in having to explain in detail how harm was caused by a specific fault or omission. They will have right of access to evidence from companies and suppliers, in cases in which high-risk AI is involved. This could mean that the businesses will have to disclose trade secrets to show that they are not liable.

Photo distributing company Getty Images are removing AI generated artworks

…Getty cites copyright liability concerns. The AI software may have been trained on scraped copyright protected images. This then has implications for the output of the AI Process. Identifying AI generated images for removal is however, not an easy process.

In the US there is a question about whether the output can be exempted from copyright infringement relying on fair use – the position would be harder for a commercial company like Getty whose business concerns dealing in images.

In the UK, the government has said that data mining will not require permission from the copyright owner by anyone with lawful access to the material even if it is protected by copyright. The aim of the policy is to make UK a location of choice for data mining and AI development.

In the EU, the data mining exceptions under EU’s Copyright Digital Single Market Directive (applicable across the EU), are quite narrow although it is possible to data mine without permission if there are no express reservations prohibiting data mining (for example coding on the webpage, contractual terms).  

OpenAI open sources Whisper – a neural network that transcribes speech in multiple languages and translates into English

…Google, Meta and Amazon all have good speech recognition systems, but Whisper sets itself apart from these because it claims to be able to recognise heavy accents, background noise and technical jargon. Having said this it is reported that it has not escaped the common pitfalls; it is less good at recognising less well represented language owing to the lack of training data and some words remain wrongly recognised.

Nvidia unveils two new large language model cloud AI services, the NVIDIA NeMo Large Language Model Service and the NVIDIA BioNeMo LLM Service

…What do these large language models do?

NeMo: Developers can customise the natural language processing models for content generation, text summarization, chatbots, code development.

BioNeMo: There are two, one for chemistry and one for biology.  Developers can customise the AI application for protein structure, and biomolecular property predictions – for drug discovery. Nvidia says ” It helps researchers identify patterns and insights in biological sequences by supporting protein, DNA, and biochemical data.”

What’s so good about Nvidia’s AI service? Developers can utilise their own training data, but it takes only “minutes to hours” not “weeks or months” to train from the start. This is because it uses “Prompt Learning” – the model is already trained using “billions of data points” – which enables the developer to adapt the AI for its use more swiftly.

Note

What is a Large Language Model?: These are machine learning algorithms that can recognize, predict, and generate human languages on the basis of very large text-based data sets.

What is the difference between Natural Language Processing and Speech Recognition: NLP is a system that can understand language, understand sentiment, understand meaning and respond to inputs or come up with further questions. Speech Recognition is the ability to interpret spoken words and convert them into text.

T-Mobile will help Pano-AI, a “disaster preparedness company” to deploy 5G connected cameras to detect wildfires in high risk areas

…the cameras will be installed in high vantage points such as cell towers. The cameras can see miles away. Pano-AI have humans in the loop, but uses AI to detect smoke as fast as possible.  

Microbots – the width of a hair could be at some point crawling inside our bodies and clean up our pipes

…the possibility is still way into the future, but scientists are now developing tiny bots that can autonomously operate. The link is to a very short YouTube clip. Take a look!

New AI uses spotted this week

Some interesting ones:

• Text to Image AI which generates new Pokemon characters (you input for example, Donald Trump – take a look, it is fun)
• AI which tracks construction site projects and identifies issues – allowing remote workers to monitor progress and identify problems early.
• Raydiant’s Digital menus at fast food places could scan you, determine your age (eg. wrinkles), mood (face scan, weather), gender, and provide you with recommendations.  

BigTech/ Data / Platforms

Content moderation

TikTok bans political fundraising

…TikTok says that it…”is an entertainment platform where people come to share their stories… including current events like elections and political issues…We …keep harmful misinformation off the platform, prohibiting political advertising, and connecting our community with authoritative information about elections. Today, we are building on that approach by making a series of changes to government, politician, and political party accounts…. This means accounts belonging to politicians and political parties will automatically have their access to advertising features turned off, which will help us more consistently enforce our existing policy….Additionally, we will be prohibiting these accounts from accessing other monetization features…. like gifting, tipping, and e-commerce, and will be ineligible for our Creator Fund.

…all in preparation of keeping out of troubles in the run up to the midterm elections.

Why is TikTok so controversial?

• Chinese owned – US citizens’ data may be accessible to the Chinese Communist party. To allay the concerns, all TikTok data has been transferred to Oracle’s data server. However, TikTok has not committed to cut off all data to China.
• TikTok can be used to brainwash users [and there are many] – by spreading propaganda by tweaking its alogorithm

TikTok will be introducing a dislike button to help Algorithms point user away from content which is inappropriate or not desired

…your dislike vote will not be shown and those that uploaded the undesired content will not be notified. This will inform TikTok about the users further, but also avoid bullying by creating numerous accounts and piling on dislikes against any particular content maker. The idea may be that content generators of bad taste end up shouting into a void unbeknownst to him or her.

You can also retract the dislike as well. It’s great – a way of content moderating your own content, and tailor things to your taste.

Inquest into the suicide of UK teenager after consuming self-harming images on Instagram and Pinterest questions social media company policies

…Instagram owner Meta’s representative said it allowed graphic pictures to be displayed to enable those in need to cry for help, and seek for community. It explained that damage could be caused by silencing those that are struggling. Instagram changed its policy to ensure that posts encouraging self-harm was removed, although it did not take down posts which admit vulnerabilities.

Pinterest admitted that its site was not safe before the teenager’s death. Meta has apologised.

The posts included close-ups of individuals cutting themselves up with a razor blade, which the coroner said was “almost impossible to watch”.

Both social media companies had recommended content about depression.

The inquest is heard as the UK considers to pass its Online Safety Bill, which includes responsibility for platforms to remove “legal but harmful” content – this includes bullying, extreme dieting, and of course self-harm. Prime Minister Liz Truss is inclined to remove such obligation (although perhaps with stricter rules for minors?), but the inquest could change her mind. A Texan appeal court has recently held that its state law which prohibits content moderation was not unconstitutional – the issue is being appealed to the US Supreme Court as the effect of the Eleventh Circuit (Florida) decision is opposite to that of the Texan Court.

Content moderation issues are incredibly difficult, and would seem to require careful evidence based policy making. However, Meta official at the inquest commented that the impact of certain material can affect people in different ways and different times and that it was really complicated to carry out research in this area.

Competition

UK Regulator OfCom to look into the competition landscape in Cloud services – Amazon, Microsoft and Google to be probed

…these hyperscalers account for about 80% of the UK cloud market. Ofcom’s aim is to identify competition issues (if any) at an early stage to prevent the market from becoming entrenched. FT reports that Ofcom is worried about barriers to market entry.  

I think they may be a bit too late. Hyperscalers had the foresight (and cash) to invest significant amounts into buying up data servers and securing data centres across the world. One benefit of this is that they have the bargaining power to purchase servers cheaply. Furthermore, they have done well to build their businesses early for at present, the world is riddled with supply chain disruption and chip shortages. Not only that, these companies have invested heavily into emerging technologies such as data processing and AI that enable them to provide tailored, useful cloud service to their customers. As a consequence, there is naturally a large moat around their cloud services. A big part of their dominance may not be due to any particular business conduct. In short, the first mover advantage may have paid off big time.

Ofcom plans to also look at how audio-visual services such as WhatsApp, FaceTime and Zoom are affecting traditional services such as telecommunication and messaging.

Connectivity

T-Mobile and Helium in partnership to provide new 5G Network

…the partnership enables subscribers to the Helium Network to have access to both Helium and T-Mobile 5G Network. Helium Network is a decentralised wireless network for IoT devices powered by cryptocurrency (Helium Network Token HNT). The network is made up of antennas owned and placed by individuals in their homes/offices (costs about $500) which can send data packets over long distances (using LoRa, or Long Range – its range is long like cellular, but data rate is lower so the infrastructure and running cost are cheaper). It is quite clever as it makes enterprising individuals help set up the network – otherwise, the cost of setting up such a network could be significant. The network’s bandwidth is hired out and owners can receive tokens for IoT businesses to use (governed by smart contracts).  One use case is the tracking of e-scooters or e-bikes which are dotted around a city. A company can track their electric scooters without having to enter into expensive cellular connectivity contracts with a network operator.

Crypto/Blockchain

The Commodity Futures Trading Commission (CFTC) enforces a DAO and its members for breaching regulatory laws

…arguably, this action puts a lot of cold water over one of the most interesting feature in web3 – DAO (Decentralised Autonomous Organisation – in short a company on a blockchain).

The charge is illegally offering leveraged and margined retail commodity transactions in digital assets using a blockchain based software protocol; engaging in activities only registered futures commission merchants (FCM) can perform; and failing to adopt a customer identification program (ie: KYC) as part of a Bank Secrecy Act compliance program.

Defendants was bZeroX and its founders. Although the bZx Protocol purported to offer users the ability to engage in these transactions in a decentralized environment—i.e., without third-party intermediaries taking custody of user assets, the company was a centralised company so the CFTC was able to enforce. Those charges are settled.

The key aspect of this development is the enforcement against the successor DAO, Ooki, apparently an unincorporated association which is carrying on using the software protocol of bZeroX. According to Coindesk interview, CFTC is enforcing against members of the DAO – making those that voted with the tokens personally responsible (and query is whether there is a distinction between the voting persons and those that are just holding the tokens). Another question which is raised is how the court papers will be served.

EVs

All Volvo’s new models to be fitted with laser sensors (LiDAR)

…these are very expensive and other auto makers have announced laser use too, but only on certain models. Volvo is known for its safety credentials, and it is thought to reinforce this reputation by its use of LiDAR. Volvo said the technology will help reduce collisions by about 9%, and accidents involving fatalities/injuries by about 20%. Volvo will also use cameras and radar systems as well.

What’s good about LiDAR

• Precise detection – it can detect “a black tyre on a black road at 120 meters ahead, or a pedestrian at 250 meters
• Can function well in darkness (unlike cameras)
• Faster detection – LiDAR works at the speed of light, radar works at the speed of sound.
• Facilitates emergency breaking capabilities which the EU is requiring all cars to have.

Volvo will purchase its LiDAR from Luminar, a company in which Volvo has a stake.

Google’s attempts at developing eVTOLs ends

… the 12 year project was called Kittyhawk, and was Google founder Larry Page’s baby. Kittyhawk did give rise to Cora, a two-seat autonomous eVTOL under a joint venture with Boeing called Wisk Aero. This JV is expected to continue, and focuses on autonomous flights, unlike other rivals.  Good to know that even Larry Page fails sometimes. He would consider that as just a step to success I’m sure.

Metaverse / VR / AR / MR / XR

Walmart land and Walmart Universe of Play enter Roblox

…Walmart hopes to attract Gen Z. It says ” Walmart Land will bring the best fashion, style, beauty and entertainment items directly to the global Roblox community of over 52 million daily users. The retailer will continue to bring the fun with Walmart’s Universe of Play – the ultimate virtual toy destination in Roblox”.

Walmart Land will have:

• Electric Island – learn to mix different beats. In October, well known artists will have concerts there.
• House of Style – virtual dressing room, strike a pose challenge

Walmart’s Universe of Play will have

• Immersive games featuring well known toys (eg. Paw Patrol)
• Rewards – collect virtual toys, earn coins and then redeem against an avatar merchandise (called verch)

Accessible on PCs, smartphones, game consoles and VR headsets.

Zepeto – Korean derived Asia’s biggest metaverse

…it is owned by Korean tech group Naver. It is reportedly a long way behind Roblox, but I would not bet against the Koreans judging by their dominance in the pop world thanks to acts like BTS, Black Pink, Super Junior (all knowledge gained thanks to my primary school aged daughter). And a lot of metaverse at the moment centers around “pop”, like gaming, fashion, cosmetics etc.

But Zepeto is an avatar centric social media platform unlike Roblox which is more a channel to chat to others whilst gaming. Unlike a normal social media platform, participants are expect to be whoever they want to be through their avatars. On the flip side you don’t really know who you are talking to (a point well illustrated in the movie Ready Player One). As you might expect from a popular metaverse it has partnerships with luxury brands and you can drive Korean cars like a Hyundai or have Korean appliances such as Samsung – to fit in your virtual home (ie: metaverse no longer a place for fashion brands only).  

Naver uses AI to identify inappropriate behaviour on its metaverse, but says policing is tricky. Brand owners may have to check out a platform’s guardrails before they do a partnership deal.   

Must check out: Nreal AR gives users an IMAX screen in the comfort of their own phone

…play whatever on your smartphone and visualise through these AR glasses. They look more like sunglasses rather than ski goggles. Nreal seems to have definitely upped the game for everyone working in AR.

Patents

Google challenges the video codec field as Project Caviar set to take off…

…the intention is to topple Dolby Vision and Dolby Atmos (also royalty free for streaming services only, but licences hardware companies and users instead) by offering purportedly royalty free formats for High Dynamic Range (HDR) Video and 3D audio.  Google can thereby avoid paying licence fees for their hardware and help increase sales of Android devices by keeping costs of hardware lower than those that opt to adopt Dolby’s video codec technology. YouTube is a must have app on devices. By having YouTube support the technology (for a more immersive experience), Google hopes to gain wide adoption both by streaming businesses and device makers.

The article notes that these technologies may in the future also contribute to VR and AR devices.

Quantum Computing

Novo Nordisk puts up $200million to create Quantum Computer especially for Life Sciences

…such device can be used for

• Formulating personal medicines, relying on analysis of genome data
• Making drug discoveries
• Modelling reactions

Semiconductors

ARM explores “strategic alliance” with Samsung

… If so in what form? An acquisition is thought to be unlikely. Nvidia bid for and failed to acquire ARM on antitrust grounds (because it would then make it very difficult for other businesses to share their IP with ARM). The same barrier is highly likely for Samsung, should it wish to acquire ARM. The great majority (estimated 90%) of today’s processors are based on ARM architecture. A strategic alliance with such a company would be more welcomed by Samsung, which has been reported that it is losing the technical edge in Smartphone Application Processors (the computer that runs the operating software, apps, processing commands, graphics, memory management etc). Chipmaker Qualcomm will be looking at the strategic alliance closely – it relies on ARM intellectual property, has been vociferous about anticompetitive effects of Nvidia/ARM acquisition, and have recently announced a multi-year strategic partnership with Samsung. ARM is also suing Qualcomm – so could ARM be trying to get some sort of leverage in the litigation by sidling up to Qualcomm’s customer?   

GM’s EV venture Cruise also to start designing chips

…buying chips of vendors too expensive so they are doing it in-house, GM says. It will also help them take control of the edge; for the edge is likely to be highly dependent on the chips and software when it comes to future cars. Which is certainly not going to be about engines. One of GM’s chips will be based on ARM architecture but it says that it is also exploring RISC-V, an open source architecture. These define the kind of software the chips can run. RISC-V can be said to be ARM’s greatest rival and threat.

Satellites/Space

NASA successfully hits an asteroid – to test our capability in averting planet killing hurtling asteroids

…the collision took place 7million miles away. SpaceX provided the engine for the satellite.

One more thing…

Ohio Jury to decide on the question of whether game maker NBA 2K infringed a tattoo artist’s copyright for depicting real life athelete’s tattoos in its games

…was the game maker’s use of the art work within the scope of US copyright fair use exemption/ de minimis use? Further, did the artist authorise the players to allow third party use of his work? Those are the questions.