Categories
AI invest crypto technology

DeepSeekers leveren paspoort in, Solana trekt anti-woke commercial terug en TikTok-VS naar Oracle?

Het belang van technologie dringt snel door in alle geledingen van de samenleving. Het is zichtbaar hoe politici worstelen met het reguleren van technologische ontwikkelingen, terwijl de doorsnee tech-CEO blijkt te beschikken over het invoelend vermogen van een aardappel.

Wie had ooit gedacht dat de Amerikaanse president zich persoonlijk zou bemoeien met de verkoop van de Amerikaanse tak van een Chinese app (TikTok), of dat de Chinese overheid de paspoorten zou laten innemen van personeel van een startup (DeepSeek) omdat hun werk van nationaal belang wordt geacht?

Er is dagelijks een nieuwe internationale rel over een techbedrijf. Alsof Elon Musk nog niet genoeg aan zijn hoofd heeft, is zijn AI-bot Grok nu verwikkeld in een controverse in India vanwege ongefilterde en provocerende reacties. Solana, een cryptobedrijf met wortels in Zwitserland en geleid door een Oekraïner en een Indiër, dacht goedkoop te kunnen scoren met een anti-woke commercial gericht op Amerika, maar moest deze bolus al snel weer inslikken.

Het is alsof de techbro’s elke ochtend opstaan met het doel opnieuw te bewijzen dat er een wereld van verschil zit tussen intelligentie en gezond verstand.

De beloning voor baanbrekend AI-onderzoek: paspoort inleveren? Beeld gemaakt met Midjourney.

Werken bij DeepSeek? Paspoort inleveren!

Het moederbedrijf van DeepSeek, het hedgefonds High-Flyer, houdt de paspoorten van belangrijke medewerkers vast, meldde The Information. Deze maatregel zou zijn genomen om te voorkomen dat medewerkers, vooral uit het R&D-team, het land en daarmee het bedrijf verlaten.

Headhunters zou ook zijn opgedragen om medewerkers van DeepSeek met rust te laten. Het toont de keerzijde aan van het belang dat de wereldmachten inmiddels hechten aan technologie, met name aan AI.

Koopt Oracle de Amerikaanse tak van TikTok?

Een verrassende ontwikkeling in de soap rond de door de Amerikaanse regering opgelegde verkoop van de Amerikaanse activiteiten van TikTok: database-gigant Oracle zou de belangrijkste rol krijgen in een herschikking van de aandelen. Bestaande Amerikaanse investeerders van ByteDance, het moederbedrijf van TikTok, zoals Susquehanna International Group en General Atlantic, zouden samen met private-equityfirma KKR de Amerikaanse activiteiten van TikTok overnemen via een nieuwe entiteit.

Oracle zou daarbij optreden als ‘vertrouwde technologiepartner’ en verantwoordelijk worden voor het beheer van de Amerikaanse gebruikersdata, waardoor de invloed van de Chinese overheid op deze gegevens wordt beperkt. Deze constructie, die het Chinese aandelenbelang in de nieuwe entiteit zou reduceren tot twintig procent, zou voldoen aan de Amerikaanse wetgeving die ByteDance verplicht om TikTok te verkopen of anders een verbod tegemoet te zien vanwege zorgen over nationale veiligheid. President Trump, eerder voorstander van een verbod, heeft nu aangegeven TikTok in de VS te willen behouden en de deadlines verlengd om tot een oplossing te komen.

Yann LeCun en Andrew Ng fileren Anthropic CEO

Mede door de groeiende afkeer van X wordt LinkedIn een steeds interessantere bron van informatie. Yann LeCun en Andrew Ng, twee grootheden in de AI-wereld, spraken zich er afgelopen week uit tegen het idee dat programmeervaardigheden binnenkort overbodig worden door de opkomst van AI. LeCun en Ng betogen dat het beheersen van programmeervaardigheden van onschatbare waarde blijft en het kortzichtig is om mensen te ontmoedigen om te leren programmeren.

Zonder diens naam te noemen spreken LeCun en Ng zich hiermee uit tegen Anthropic CEO Dario Amodei, die dacht ook eens dapper te moeten doen in zijn PR-strijd tegen OpenAI topman Sam Altman, en zich liet verleiden tot de uitspraak dat binnen een jaar negentig procent van al het programmeerwerk wordt gedaan door AI. Laten we een gokje wagen: in Amodei’s wereldbeeld moet dat gaan gebeuren met software van zijn Anthropic.

LeCun reageerde kort maar krachtig:

Leer wiskunde, natuurkunde, techniek en informatica. We krijgen misschien binnenkort superintelligente AI-assistenten, maar wij zullen hun baas zijn. We moeten de basis begrijpen om hen te kunnen leiden.”

Dario Amodei’s goedkope PR-teksten

Amodei probeert continu het nieuws te halen met slappe PR-teksten. Zo zei hij onlangs dat AI binnenkort ‘een land van genieën in een datacenter’ zou genereren. Om deze kreet te pareren nam LeCun meer tijd:

“Door simpelweg LLM’s op te schalen, gaan we geen AI op menselijk niveau bereiken. En dat gaat gewoon niet gebeuren. Er is geen enkele manier, oké, absoluut geen manier. En wat je ook hoort van sommige van mijn avontuurlijkere collega’s, het gaat niet gebeuren binnen de komende twee jaar. Absoluut geen enkele kans. Het idee dat we een ‘land van genieën in een datacenter’ zouden hebben – dat is complete onzin. Dat gaat absoluut niet gebeuren.
Wat we misschien gaan hebben, zijn systemen die zijn getraind op voldoende grote hoeveelheden data, zodat elke vraag die een redelijk persoon zou kunnen stellen, een antwoord kan vinden via die systemen. Je zou het gevoel kunnen hebben dat je een PhD naast je hebt staan. Maar het is geen PhD die je naast je hebt, het is een systeem met een gigantisch geheugen en terugvindcapaciteit. Geen systeem dat oplossingen kan bedenken voor nieuwe problemen.’

– Yann LeCun

LeCun verkondigt al langer dat het opschalen van LLM’s, zoals ChatGPT van OpenAI, Anthropic’s Claude, of Elon Musk’s Grok, niet zal leiden tot AI die mensachtige intelligentie bereikt. Volgens hem kan tekstdata alleen geen menselijk begrip opleveren. LeCun stelt niet geïnteresseerd te zijn in LLM’s om mensachtige AI te creëren en denkt dat grote technische en architecturale doorbraken nodig zijn om AI in het huidige groeitempo voort te zetten. Hij ziet geen toekomst in puur schaalbare LLM’s om menselijke intelligentie te evenaren.

John Battelle nuchter over AI-agents

Journalist en ondernemer John Battelle, vooral bekend als mede-oprichter van Wired en The Industry Standard, betoogt dat volledig autonome AI-agenten voorlopig onhaalbaar zijn om een andere reden dan technologische beperkingen. Hoewel de technologie vooruitgaat, wijzen praktische obstakels zoals databeheer, trainingskosten en bruikbaarheidsproblemen erop dat echt autonome AI-assistenten voorlopig een droom blijven.

Zelfs het gerucht dat OpenAI van plan is om AI-agenten aan te bieden voor $20.000 per maand lijkt eerder een ambitie dan een realiteit, gezien de enorme operationele uitdagingen. De kern van het probleem zit in het delen van de benodigde data om autonome AI-agenten taken te laten uitvoeren.

Anders gezegd, er is overal data, data, data, maar zonder connectiviteit is er geen druppel te drinken. Wil je dat je slimme nieuwe AI-agent een vlucht voor je boekt? Dan moet die samenwerken met, laten we eens kijken… alle grote luchtvaartmaatschappijen, alle belangrijke betalingssysteem, alle grote reisplatforms, alle grote agenda-applicaties en… nou ja, dat is al genoeg om de gemiddelde AI-ontwikkelaar te laten struikelen. 

Voor elke mogelijke situatie zou een ontwikkelaar een aangepaste programmeerinterface moeten coderen, om nog maar te zwijgen over het feit dat hun zakelijke collega’s met elk bedrijf een deal moeten sluiten om toegang tot de data te krijgen. Zulke obstakels zijn voor de meeste startups vrijwel onmogelijk te overwinnen. Dit is de reden waarom je op dit moment geen AI-agent hebt die veel voor je doet – en dat zal voorlopig ook niet gebeuren.

Maar het belangrijkste zijn de overkoepelende bedrijfsmodellen van de grote technologiebedrijven, en in het huidige technologielandschap is het bedrijfsmodel bepalend voor de toekomst. Hoe zou onze vluchtboekende AI-agent de advertentie-inkomsten van Google Flights beïnvloeden, om nog maar te zwijgen van de zoekresultaten? Hoe zou het de betrokkenheid en commerciële inkomsten op platforms zoals Instagram of WhatsApp beïnvloeden? Zou onze hardwerkende AI-agent de schijnheiligheid van Apple rondom consumentenprivacy blootleggen – een schijnheiligheid die de hele marketingstrategie voor iOS en de iPhone drijft? 

Ik zou nog wel even door kunnen gaan, maar ik denk dat je het punt begrijpt: Geconfronteerd met de onzekerheden die zulke agents met zich meebrengen, zullen deze grote bedrijven waarschijnlijk niet in actie komen. En als ze ooit een AI-agent ontwikkelen die onze eenvoudige visie kan uitvoeren, kunnen we van één ding zeker zijn: die agent zal niet voor ons werken. Hij zal voor hen werken – en daardoor falen als consumentenproduct.”

CoreWeave succesvol met eerste ‘Pure-Play AI IPO’

De eerste ‘AI cloud services IPO’ lijkt een groot succes te worden. Volgens de laatste schattingen gaat CoreWeave, dat wordt gesteund door aandeelhouder Nvidia, richting de $3 miljard ophalen op een waardering van $32 miljard.

CoreWeave bouwt GPU-aangedreven cloudinfrastructuur, speciaal ontworpen voor AI-toepassingen. In 2017 opgericht door drie beurshandelaren als crypto-miningbedrijf, werd het bedrijf na een correctie in de cryptomarkt omgedoopt tot CoreWeave. Ze haalden geld op en kochten zoveel mogelijk GPU’s van failliete cryptominers in de overtuiging dat deze chips de ontwikkeling van AI zouden aandrijven. Hun gelijk werd bewezen toen in 2022 ChatGPT op de markt kwam.

De beursgang van Coreweave wordt gezien als een test voor het vertrouwen van beleggers in AI-infrastructuurbedrijven. Gezien dit succes zullen er snel meer IPO’s volgen door allerlei bedrijven die zich zullen vermommen als ‘AI-infrastructuur.’

De beloning voor een overname van $32 miljard: -1.30%

Google koopt Wiz voor $32 miljard, boeit niemand

Google deed de grootste overname in zijn geschiedenis met de aankoop van cloudbeveiligingsbedrijf Wiz, maar de reactie vanuit de tech- en financiële wereld bleef uit. De overname had nauwelijks invloed op de aandelenkoers van Google, vooral in vergelijking met concurrenten zoals Microsoft en Amazon. Beleggers twijfelen hoeveel de overname daadwerkelijk bijdraagt aan Google’s positie in cloudbeveiliging. Google blijft na Amazon en Microsoft de derde speler in clouddiensten.

Elon Musk’s Grok sensatie in India 

Het begon met een vraag op X aan Grok, dat met enige vertraging antwoorde, inclusief beledigingen in het Hindi. Musk positioneerde Grok immers als een ‘anti-woke’ chatbot met een ongefilterde, speelse toon, geïnspireerd door het boek The Hitchhiker’s Guide to the Galaxy. In dit geval mengde de politie van Delhi zich zelfs in de interactie met de bot, wat de populariteit verder aanwakkerde.

Vooral de manier waarop Grok zich mengt in politieke discussies in India is ongebruikelijk. De chatbot uitte kritiek op premier Modi en noemde oppositieleider Rahul Gandhi ‘eerlijker en beter opgeleid dan Modi.’ Critici van Modi en voorstanders van vrije meningsuiting zien Grok als een frisse wind in een tijd waarin de vrijheid van meningsuiting onder druk staat, anderen waarschuwen voor ongefilterde verspreiding van leugens.

Boek over Facebook verschrikkingen bestseller

Vorige week was al duidelijk dat de rechtzaak van Meta tegen Sarah Wynn-Williams om de verspreiding van haar boek over haar tijd bij Meta/Facebook te verbieden, tot grote problemen voor Zuckerberg zou leiden. Door alle publiciteit kwam het boek op nummer 1 binnen in de bestsellerlijst van de New York Times voor non-fictieboeken. Zuck bewijst hiermee opnieuw dat hij het verschil nog steeds niet kent tussen intelligent en slim.

Solana verwart crypto met… genders?

Het leek deze week een wedstrijd in stupiditeit tussen Meta en Solana. Het crypto-platform bracht een commercial uit met een anti-woke boodschap die voor zoveel controverse zorgde dat Solana zich gedwongen voelde hem na een paar uur te verwijderen. De commercial was juist bedoeld om Solana neer te zetten als een nuchtere, innovatiegerichte blockchain.

Anatoly Yakovenko, medeoprichter van Solana, bood nog excuses aan, maar het kwaad was al geschied. De vraag is hoe een bedrijf het verzint om een commercial uit te brengen waarin een man, die Amerika moet voorstellen, tegenover een therapeute op de bank zit en teksten uitkraamt als ‘ik wil technologie uitvinden, geen genders.’ Het is alsof Ronald McDonald bij de juf klaagt dat hij geen staatsleningen wil verkopen; het slaat nergens op.

In aflevering 8 van de NFA Podcast bespreken Nish en ik verder de onverwachte concurrentie tussen Pump.fun en Raydium, Dubai’s proefproject voor de ’tokenization’ van vastgoed en het gebrek aan grote Indiase technologiebedrijven, vergeleken met Chinese giganten zoals Tencent en Alibaba.

Aflevering 8 van de NFA Podcast met Nish & Frackers is nu beschikbaar:

Bedankt voor de belangstelling, tot volgende week!

Categories
technology

Gamble: TikTok only goes black temporarily in the U.S.

 Mark Zuckerberg's dream comes true: his major competitor TikTok no longer works in the U.S.

The new year has barely begun and, especially on social media, there is already a lot going on. Mark Zuckerberg declared in a video dressed as some kind of pimped-up Joardy that racists and other goofballs are allowed to go wild on his Facebook and Instagram as usual. Of course, this is to please the MAGA universe, because anything goes to sell ads in the world of Zuckerberg and his famous rubber backbone.

TikTok no longer works in the U.S.

In the US today, TikTok is no longer operating, although Trump has signaled tomorrow on his first day in office that he will give the popular app 90 days to comply and sell TikTok's US operations.

From the fact that the CEO of TikTok was invited to Trump's inauguration, it can be inferred that a permanent ban on TikTok in the U.S. will not be too bad. It is more likely that Trump will force parent company ByteDance to sell TikTok, for example to his new hypeboy Elon Musk. It is highly questionable whether Trump will actually start a trade war with China, the country that is a crucial market for American giants such as Tesla from friend Musk, as well as Apple, Intel and Boeing.  

Categories
technology

Apple Vision Pro better than expected

It costs a little, but then you have something.

Apple has since been surpassed by Microsoft as the world's most valuable company, and the former stock market darling still got a whirl from Wall Street despite rising sales, while virtually all tech companies rose. Perhaps that is precisely why there was a lot of attention on the launch of the Apple Vision Pro, the mixed reality headset that Apple itself interestingly describes as "spatial computing.

When the Apple Vision Pro was announced last year, I wrote

'All the omens are that the Apple Vision Pro will be a flop - a flop by Apple standards, that is. But that's not a bad thing at all. At least Apple is trying to develop something new again, and that's better than unimaginatively buying back its own shares for hundreds of billions, as it has in recent years.' 

Because the price is too high at $3,500 to break open a mass market, there is no reason to change opinion about the Vision Pro's short-term business impact.

Apple is on its way to $500 billion in annual sales, so before any new product raises an eyebrow when going through the annual figures, it has to come close to the annual sales of Apple's least contributing product. That's the iPad, which still did $7 billion in sales last quarter. To get anywhere near that, Apple would have to sell a few million copies of the Apple Vision Pro, which is not going to happen with the current model at this price.

Vanity Fair was invited by Apple CEO Tim Cook to learn about the Apple Vision Pro, which led to this revelation from the reporter:

'When I turn it off, every other device feels flat and boring: my 75-inch OLED TV feels like a TV from the '90s; my iPhone feels like a flip phone from yesteryear, and even the real world around me feels surprisingly flat. And here's the problem. 

In the same way I can't imagine driving a car without a stereo, in the same way I can't imagine not having a phone to communicate with people or take pictures of my children, in the same way I can't imagine trying to work without a computer, I can envision a day when we all can't imagine living without augmented reality (AR). 

When we become more and more encapsulated by technology, to the point that we crave these glasses like a drug [...], the dopamine rush that this resolution of AR can deliver.'

Most reviews were less lyrical than this one, but mostly positive. The bottom line is that Apple has once again succeeded in developing a surprisingly special and high-quality product. And yet, there's something nagging.

Apple tries to solve an unsolvable problem

Wired correctly states that a "killer app" has not yet been found for the Apple Vision Pro. It is not yet the ultimate entertainment device and that is not because of the quality of the image, the sound or the controls, because they are extremely good. It's because of the applications, and then not even the "content," the traditional video narration form in picture and sound. The problem lies in the lack of new communication applications between people.

Now I am not neutral when it comes to VR and AR, having worked at VR pioneer Jaunt for a few years. I experienced the exact same experience in Jaunt's test lab as the Vanity Fair journalist, because good VR has an almost hallucinatory effect. But you remain a spectator in someone else's film.

And the core of the Internet's success is not information, transaction or entertainment. It is communication between people. The great breakthrough of social media was not caused by expensive content from movie studios or game developers, but by movies like Charlie Bit My Finger.

Despite all the success of social media like Facebook and Instagram, the messaging service Whatsapp is being used more intensively by users. And just when it seemed that the market for messaging apps was saturated, Telegram managed to attract as many as a quarter of a billion new users in 2023, bringing the total number of users to 700 million people. The demand for communication options between people seems inexhaustible.

So the big question for Apple becomes not how it can develop even flashier VR and AR applications, or how it gets Netflix to create apps for the Vision Pro; but whether it manages to develop interpersonal communication applications for the Apple Vision Pro that are as useful, funny and addictive as ever text messaging. As an enthusiast, I wish Apple would focus on that and, for example, permanently disband its entire automotive division. How many electric automakers does the world need?

Is TikTok the answer?

Especially when it comes to communication between people, TikTok has proven to be a phenomenon. When it seemed like the social media market had been completely nailed shut by Facebook, Twitter and Instagram, with Snapchat and Twitch as boutique stores, dances appeared on this originally Chinese app that were emulated worldwide. For dance requires no spoken language, only a sense of rhythm or a glaring lack of embarrassment.

Meanwhile, TikTok has become so big that Wired wrote an extensive profile on the company's Singaporean CEO, who had to answer to the US Congress for the second time last week, with a senator going out of his way to appear as racist and anti-Chinese as possible to his constituents. Incidentally, the Singaporean Internet responded within 24 hours with a hilarious video.

I'm curious what a TikTok app on the Vision Pro would look like and what you could do with it. Dance together, or watch movies together, so that using the Vision Pro at least becomes a shared experience?

Or is it Joe Rogan?

Once upon a time, the world's most popular podcast maker Joe Rogan hosted the TV show Fear Factor, a derivative of Now or Neverland. In that program, contestants from the Netherlands and Belgium had to complete tasks such as jumping out of a building while holding an egg that was not supposed to break, or eating worms while Hans Kraay Junior yelled at them "do it for your country, eat those worms for the Netherlands!

Joe Rogan, the American Hans Kraay Junior, signed a new contract with Spotify this week that will net him as much as a quarter of a billion dollars. Interestingly, it is not even an exclusive contract with Spotify, so Rogan will be seen and heard on multiple platforms.

Rogan's podcasts are recorded representations of the most basic form of communication since the dawn of mankind: two people talking to each other. Rogan's success lies in his curiosity.

He is actually interested in his guests and never tends to want to be clever at the expense of his guests. Maybe he's not that smart, which is always the criticism of him, but perhaps that's exactly what makes his podcasts accessible to a wide audience.

I would not be surprised if there are millions of people who, with an Apple Vision Pro on their heads, want the feeling of sitting at the table next to Joe Rogan and Elon Musk, or Quentin Tarantino or Lance Armstrong, as a third person. Not even to participate on equal footing, but to experience an interesting conversation up close. The mere fact that this kind of application is relatively easy to make is a reason to conclude that the Vision Pro is underrated.

Because it may quietly take five years and three versions of the Vision Pro before the device finds its killer apps combined with a good price, but then Apple will have a new successful form of personal computer on its hands alongside the Mac, iPhone, iPad and Apple Watch. Losing seventy billion in market cap the week the Apple Vision Pro hit the market? Investors should be ashamed of themselves.

Categories
technology

Hamas gets a helping hand from social media

It is not possible to write cheerfully about new innovations when the world is on fire. Especially not when the world's most widespread innovation of this century, social media, is being used by terrorists and imbeciles to spread hatred and prevent peace.

What is the most important story on Meta's news page?

Historian Yuval Noah Harari aptly summed up the misery on CNN: the goal of Hamas and related clubs is simple, namely, "to assasinate any chance of peace." Spreading their atrocities through social media is a deliberate tactic in doing so. The hatred and disgust evoked by the images will eradicate any sense of compassion, let alone willingness to compromise. Exactly the intention of Hamas.

It proved virtually impossible on Facebook, Instagram, TikTok, X and Youtube last week to avoid the topics of Israel and the Gaza Strip. This seems logical at first, were it not for the fact that I was continuously served messages from all sorts of people I do not follow at all. No matter what I clicked away and pushed away, the distasteful nonsense from Gaza experts, many of whom were self-proclaimed AI experts until last week and NFT insiders last year, kept flooding my timeline.

Facebook and TikTok respond only after warnings

It took until Friday for Western governments to issue warnings. The EU ordered Meta to do something about all the disinformation being poured out over the world via Facebook and Instagram. Zuckerberg, meanwhile, understands that you have to move then, just as someone running a red light always pretends to cross as fast as possible. TikTok received a similar warning, as did YouTube.

The statement Meta issued is so saltless and heartless that it must have been written by a bad AI application. But look especially at where Meta places that statement: small in the right-hand column. The most important place on a website, big in the middle with animated visual, is still dedicated to the boring chatbots that were already announced on Sept. 27.

Next time a Facebook livestream by Hamas terrorists wearing smart Ray Bans?

Zuckerberg was able to chat with Lex Fridman in a promotional podcast about his Metaverse efforts last month, but where was he last week when terrorists broadcast child murders over his networks? Sure, it's brave for a nerd to spar with professional cage fighters, but it takes real courage to go live on CNN to talk to Christiane Amanpour about your contribution to the spread of terrorist outrages.

Meta's news page pays more attention to the new Meta Smart Ray Bans, which allow live streaming from a camera in those glasses on Facebook and Instagram, than to what Meta will do against abuse by groups like Hamas.

If Zuckerberg doesn't intervene, the wait is on for a Hamas idiot (fighter is the wrong word for someone who kidnaps a girl at a music festival) with a fashionable 'Rebel Black Headliner Ray-Ban Meta' smart glass on his head committing mass murder; because it's easier to operate your Kalashnikov when you have your hands free while going live on Insta.

The business model of social media is free popular content

It's crude, but at its core, Hamas delivers what social media loves: lots of viewed free content. Meta's business model is to sell advertising through Facebook and Instagram to people who watch content for which Meta pays nothing. That's much more lucrative than all that cumbersome stuff from Disney or Netflix, who have to create expensive content.

At its core, Meta doesn't care what that content is, as long as they don't get hassled about it by advertisers. As Zuckerberg likes to say, "move fast and break things. The world has been able to see how that works out.

The problem is not new

Once a young Internet entrepreneur said:

"We are largely similar to the phone company. We only transmit information; we don't have a say about what is transmitted. We only select a little bit: if we passed on everything on the Internet, our computers would crash. And we also rejected a discussion group that did spread child pornography very clearly. But we could only tell that from the name of that group, because it would be impossible to read everything said in it."

That young Internet entrepreneur, that was me. And I said this in newspaper Trouw in November 1995, almost 30 years ago. I was 27 at the time and I sincerely believed that the operator of a network had no business meddling in the content that people transmitted over this network.

Not long after, several of our subscribers began distributing child pornography through newsgroups devoted to innocent topics such as bird watching or stamp collecting.

We had worked for years as communications science graduates to build an Internet provider that would provide Internet access to everyone in the Netherlands, allowing people to communicate and gain knowledge, but a certain group of subscribers decided to spread the worst imaginable misery through our network.

Filter freedom of speech

In all civilized countries there is a limit to freedom of speech somewhere. I didn't know then and I don't know now exactly where that line is, but we all know when it is crossed by miles.

We actively reported our own child pornography distributing subscribers to the vice police, which led to several criminal cases. It was the darkest period of my time as an entrepreneur and they were the worst things I have ever seen. I was deeply impressed by the commitment and professionalism the police showed in these cases. The detectives always looked immediately to see if the images were already known, so that hopefully there were no new victims. Fortunately, this was true in all cases.

With pain in our hearts, we closed access to many newsgroups, thus limiting both subscribers' access to information and their ability to disseminate information. The good people suffered from the bad people, because 99% of the messages could be classified as normal communications.

Therefore, I can well understand that Mark Zuckerberg and other entrepreneurs and executives who have worked hard to build their networks find it difficult to now have to limit and control what they have built. But global reach and influence comes with commensurate responsibility, to which Zuckerberg in particular seems blind.

No new glasses, but working filters

Meta, X, TikTok and YouTube should now not focus at all on developing technology that will make it even easier for everyone to distribute even more content, such as via an avatar in the Metaverse or with smart glasses with built-in cameras; the focus should be on regulating which users broadcast what content over their networks and who has access to viewing that content.

The same kind of algorithms that YouTube and Facebook use to see what content is going viral and promote it faster can also be used to estimate what content needs to be looked at more closely for objectionable substance. Good use of the lightning-fast developments in AI and text and image recognition should be able to limit spread of much misery.

Meta's market cap is over $800 billion and the company makes tens of billions in profits. That's more than enough budget to build functioning systems. The same is true for YouTube and TikTok, while the ailing X is surely able to get a loan from its owner Elon Musk.

Over the past week, children around the world have seen images on Instagram and TikTok that no one should ever see. Preventing that misery is what Zuckerberg, Musk and the other chiefs of social media should focus on.

Four years ago, Zuckerberg had a long conversation with Yuval Noah Harari. It would be good for them to talk again now. The value of democracy is at stake.If America (Meta, YouTube, X) has moral superiority over China (TikTok), now is the time to prove it.

Categories
AI technology

TikTok loses six billion dollars and according to ChatGPT, I'm a Pussycat Doll

Never trust information from AI platforms, because even though it is my dream: I am in this picture, but I am not (yet) a Pussycat Doll, despite the striking resemblance.

TikTok loses six billion sales due to Indonesian web shop ban

This week's most notable news was not the lawsuit against Sam Bankman-Fried, because that curly clearly stole as much money as he could from his customers seems clear, but the Indonesian government' s ban on TikTok's webshop. And according to ChatGPT, I am finally a member of the girl group The Pussycat Dolls.

Southeast Asia, an internationally still underrated market with over 675 million people, about half more than the EU, is one of the largest markets for TikTok with over 325 million users per month.

Indonesia has as many as 125 million TikTok users among its 278 million population. That includes six million sellers and millions more creators who make money by using TikTok Shop to promote products. TikTok is much more than a social media platform in Asia; it is an economic force.

Online retail in Indonesia has surged in recent years. The value of e-commerce sales has increased sixfold since 2018 and is estimated to reach $44 billion next year, according to the central bank.

While Indonesia’s e-commerce market is dominated by Shopee, Tokopedia, and Alibaba-backed Lazada, TikTok Shop had made significant inroads since launching in April 2021 and was reportedly on track to handle as much as six billion dollars in transactions in Indonesia this year, on which TikTok earns 5% commission, a whopping three hundred million dollars.

Until last Thursday afternoon, when TikTok had to shut down its web shop because the Indonesian government had given it an ultimatum last week. President Joko Widodo had previously indicated that TikTok's influence on Indonesia's economy was out of control and local retailers simply could not compete against TikTok.

Whereas the appearance of TikTok's CEO in the U.S. Congress still led to media reports worldwide, this form of government intervention went unnoticed by most of the world.

Anthropic wants to raise another $2 billion, now from Google again

Just last week I wrote about the four billion dollars Amazon is putting into Anthropic, the major competitor of OpenAI (maker of ChatGPT, about which more later), and this week it appears that Anthropic needs even more money: as much as two billion dollars.

The striking thing is that Anthropic wants to raise this money in part from Google, which previously invested three hundred million dollars in the company for a ten percent stake. Now Google would have to pay about tenfold for the same percentage. And this is especially painful because of what leaked out about the problems Google is having operationally to keep Anthropic, maker of Claude, up and running.

Google's loss is Amazon's gain

Last month, according to this article, a team of 50 people at Google Cloud worked weekends to fix an unstable Nvidia cluster that Anthropic has running there:

"To fix the faulty part of its service - an underperforming and unstable Nvidia H100 cluster - Google Cloud leadership initiated a seven-day-a-week sprint for the next month. The downside of not making it work, the senior engineer said, was "too great, especially for Anthropic, for Google Cloud and for Google."

Then this team was told that much earlier it had already been arranged that Amazon would invest four billion in Anthropic and that Amazon's cloud service AWS would become the "primary cloud provider for mission-critical workloads." In other words, Amazon would do all serious work and heavy lifting for Anthropic, and only some fringe stuff would remain with Google.

To top that off the news came this week that Google is a candidate to invest another two billion dollars in Anthropic. No one has to wonder what those seven-day-a-week sprinting Google engineers think about this.

Anthropic helps FTX creditors

Those watching with delight the immense valuations of the new generation of AI companies are the creditors of FTX that previously invested $500 million in Anthropic. Their hope is that the stake FTX holds in Anthropic will eventually offset much of the losses they have suffered so far from all the misery and theft at FTX.

Evil applications of technology

There was more very bad news on the technology front. First, a Replika chatbot appears to have encouraged a deranged man to assassinate Queen Elizabeth. It remains amazing that developers fail to see how their well-intentioned technology can be used by criminal or spineless characters.

A confused mind in the UK felt empowered by this kind of message from a chatbot

Then 23andMe, the U.S. biotechnology and genomics company that offers genetic testing services to customers who send a saliva sample to their labs and get back a pedigree and genetic report. The company confirmed this week that customer profiles offered on the Internet for $1 to $10 per profile actually came from their site.

It is likely that previously used passwords were used and 23andMe was not hacked, but factually that is irrelevant. Ars Technica aptly summarizes:

'On Friday, The Record and Bleeping Computer reported that one leaked database contained information for 1 million users of Ashkenazi descent, all of whom had opted into the DNA matching service. The Record said a second database contained 300,000 users of Chinese descent who had also opted out.

While there are benefits to storing genetic information online so people can trace their origins and locate relatives, there are obvious privacy risks. Even if a user chooses a strong password and uses two-factor authentication, as 23andMe has long recommended, their data can still be collected in scraping incidents as it recently confirmed. The only sure way to protect it from online theft is to not store it there in the first place."

So the hackers had a specific preference for selling data of Jewish and Chinese people. But this kind of data simply should not be stored online at all. Period.

Spotlight 9: focus on Canva

Tesla is the winner of the week, Ethereum the loser.

It was a positive week in the stock markets with Tesla passing Meta (Facebook) in market value, but I want to talk about a company that has a great chance of a successful IPO or a huge hit in a sale: Canva.

Forbes Australia has an excellent profile on this Melanie Perkins-led Australian-origin company, which this week launched the AI-powered Magic Studio.

Not long ago, InDesign was the software package that only professionals could use to create things, which any simple soul can now create with Canva. The proof? The above graphic, beautiful in ugliness but functional, I have been creating with Canva since the beginning of this newsletter.

There's even a spreadsheet behind it that Canva creates, in which I only have to enter Friday's closing prices. And just now I did some tinkering with the new feature that lets you simply generate videos based on text and images. What I make doesn't look like anything remotely acceptable yet, but I think this will be a very popular application.

Here is a brief look at the company's recent performance:

  • valuation: Canva is now valued by investors at nearly $40 billion.
  • revenue: The company is on track to achieve annual sales of $1.7 billion.
  • users: Canva has 150 million monthly users, of which 16 million are paying subscribers.
  • Profitable: Canva is profitable this year for the seventh consecutive year.

Canva is so clever, creatively, technically and financially, that Adobe or Microsoft is going to buy it for at least $50 billion. Adobe must hurry, or else it will be unable to pay Canva at all and could even be replaced by Canva.

Microsoft is the most natural buyer because Canva is the most fitting, Web-based extension of the Office suite. And Microsoft CEO Satya Nadella buys companies like a Frenchman buys croissants.

Quick takes on other news

David Perell had a long conversation with Marc Andreessen

The link goes to a tweet in which Perell summarizes the 19 key points from the conversation, and the full video is here. I have been a fan of Andreessen (creator of the first Web browser and top investor) for nearly 30 years because he continues to couple a tremendous wealth of knowledge and experience with curiosity and enthusiasm.

New General Motors CTO leaves after one month

You rarely hear this: someone who has worked at a company for years gets a new position there, gives an interview and the following week he is gone. Ex-CTO Gil Golan wanted GM to build its own batteries and that turned out to be a dead end for him, excuse the pun.

Nvidia competes with its own customers

Nvidia has its own cloud service, DGX Cloud, with servers located in the data centers of Microsoft, Oracle and Google. The Information reported this week that Nvidia is looking into building its own data centers, which would put it in direct competition with its biggest customers, who in turn are diligently trying to develop their own AI chips, Nvidia's core business. I've been trying to make a chart of the AI playing field with Canva for weeks, but when everyone in the space starts tinkering with each other's business, it gets tricky.

Rumor: Apple wants to buy Formula One broadcast rights

Because of Apple' s success with the MLS broadcast rights since Lionel Messi has been playing there, Apple is rumoured to be willing to pay  as much as $2 billion a year for Formula One's worldwide broadcast rights. This would be a gradual transition because there are still many long-term contracts with various broadcast rights holders around the world.

Business Insider tested ChatGPT's new features

An excellent overview of ChatGPT's new features, but I was particularly interested in the option that you can upload a photo and ask questions about it. To make it difficult for ChatGPT, I chose a photo from the summer of 2007 when I was trying to get ahead of a midlife crisis by scoring a late summer hit written by Arjen Lubach with a flower grower friend disguised as Mexican rappers.

I expected all kinds of replies, but not that ChatGPT  would conclude that we were members of American girl group The Pussycat Dolls. (It doesn't mean I'm not flattered.) In short, even if OpenAI would be valued north of$100 billion because of ChatGPT ; there is still a lot to improve in the world of AI.

Categories
NFTs

Zuckerberg's midlife crisis with NFTs on Instagram

Facebook, sorry, Meta CEO Mark Zuckerberg announced yesterday that in the coming months it will become possible to mine and trade NFTs on Instagram. It is an attempt by Zuckerberg to stay relevant to young people and appease the stock markets, dwarfed by TikTok on the one hand and, on the other, the huge market value of NFT marketplaces like OpenSea.

Engadget reports this about the session with Zuckerberg at the leading festival SXSW:

"We're working on bringing NFTs to Instagram in the near term," he said. He didn't detail exactly how that would take shape, but suggested people would be able to show off their existing NFTs and potentially mint new ones. "I'm not ready to kind of announce exactly what that's going to be today. But over the next several months, the ability to bring some of your NFTs in, hopefully over time be able to mint things within that environment."

It was never a poet.

SXSW has increasingly changed from the springboard where revolutionary tech talents take the axe to the carrot of ruling powers such as the Facebook empire, into a kind of bar-dancing where mom and dad also have a dance after Friday afternoon bingo. That can be a lot of fun, and Uncle Frackers himself has organized regular dings during SXSW. But it was no coincidence that the session with Zuckerberg, who likes to present himself as a great technological visionary, was also open to convention attendees with a film and music badge, with the recommended knowledge level: "beginner.

Level: Beginner

Meta's biggest money-maker was always Facebook, which has now turned into a kind of online SBS6 for people over 50. Instagram is also moving hard in that direction, while Whatsapp remains difficult to monetize. The playful vistas Zuckerberg paints about the Metaverse have already spooked investors: this year, FB (short for Meta, cough) shares fell as much as 43%. That came after Meta estimated revenue figures for 2022 much lower than the market expected, primarily a result of Apple's tightened privacy measures. In short, Apple is making it harder for Meta to track and stalk you everywhere online.

With that have come some new problems for Zuckerberg. First, it is clear that his weapon in the battle for youth, Instagram, is being completely outpaced by TikTok. Every successful feature of TikTok is copied by Instagram while it itself hardly introduces any innovations that TikTok adopts. Instagram also turns out not to yield good conversions for advertisers, in other words: nice for branding like a billboard along a busy highway, but not a marketplace where people make transactions on which Zuckerberg pockets a percentage each time.

But there is a fast-growing marketplace achieving huge revenues and good margins: OpenSea, the marketplace for NFTs that was recently valued at over $13 billion. Zuckerberg has obviously followed the rise of TikTok and OpenSea closely and hopes to kill two birds with one stone with NFTs on Instagram. That desire is not surprising, as young TikTok users are largely crypto users and the primary target audience for NFTs. TikTok is a meeting place for the digital vanguard, just as OpenSea is for traders in NFTs.  

Zuckerberg's announcement about NFTs on Instagram is very reminiscent of the puffed-up marketing copy he churned out about Libra. That was that crypto currency from Meta that was going to be the whole thing and died a quiet death within three years. For as unimaginably long as it seems, Libra was only announced three years ago, in 2019, and has already been buried without fanfare.

It's not to say that NFTs on Instagram are definitely going to fail, just that all the conditions for total failure are there: a poorly converting medium for transactions, an aging target audience that comes for very different things than NFTs, and stronger competitors on the chessboards where it counts: the active youth at TikTok and the NFT merchants at OpenSea. The wait is for TikTok to do something with NFTs. Perhaps in a partnership with OpenSea, the dream scenario for traders who go short on Meta.

No, the more important crypto and NFT news of the past few days was elsewhere than Meta. Singapore announced it was going to impose income tax on NFTs, very pragmatic and logical. It shows how big NFTs already are in Southeast Asia. And Consensys, maker of tools like Metamask, raised $450 million from investors. Now if Consensys would use those tokens to develop user-friendly crypto wallets, the breakthrough of crypto and NFTs to the general public is near.