Alexander Chalkidis

Tag: ai

  • 3 reasons ChatGPT has no chance against Google

    3 reasons ChatGPT has no chance against Google

    I had to laugh out loud recently when I read about the legal case against Anthropic. It was about the fact that they used books to train their AI. Google started its book scanning project, codenamed “Project Ocean,” in 2002. This project aimed to digitize a vast number of books and create a massive digital library. In 2004, Google officially announced its Library Project, partnering with major research libraries like Harvard, Stanford, Oxford, the University of Michigan, and the New York Public Library to scan books from their collections.

    The project grew quickly. By 2005, Google had already scanned more than 10 million books. The project has faced some legal challenges, particularly concerning copyright, including a class-action lawsuit from the Authors Guild of America, which Google ultimately won. As of October 2019, Google announced it had scanned over 40 million titles. The company has a long-term goal of scanning all 130 million distinct books estimated to exist in the world.

    That’s the first reason nobody comes close to Google for sheer input data. They set out to “digitize the planet’s data” a long time ago and they have data Open AI can’t even imagine getting. Ever.

    The second reason is efficiency. People love talking about how power hungry AI is. Guess what? Google has been optimizing for decades now. It is not a one off that Open AI asked them to use their infrastructure. I remember maybe 20 years ago I ran into some friends of friends in Holland. Database experts, they had developed a little trick to make things slightly more efficient and…Google had bought it instantly.

    Unless you are some iPhone toting schmuck that only uses Chat GPT you know this already. Gemini is just so much faster. And they keep lowering the cost per query. GPT-4 is estimated to be around 0.42 Wh per query, while a long prompt can be much higher, sometimes exceeding 33 Wh. Google Gemini 2.0 Flash is estimated to be highly efficient, with energy use as low as 0.022 Wh per query. Claude 3.7 Sonnet ranks high in eco-efficiency among its peers at 0.81 Wh. (All these figures with massive pinches of salt but you get the picture.) And then there is the energy cost of training of course where again Google is years ahead because, well, Google has been doing AI for years. (Auto replies in Gmail , Google Photos, etc.)

    The third reason Open AI can’t catch Google is our fav, the network effect. The minute they switched all Google searches to become AI searches it was clear. They are not messing around. Not by a long stretch can any other company come close to the amount of data Google has on what people want to know. And they have been optimizing the answers they provide for decades. They were probably laughing at the recent user negativity to Chat GPT 5. Google has dealt with stuff like this since it’s inception, always experimenting with ways to present results.

    So will Open AI die? Of course not! Google knows how to play the anti monopoly game. Open AI will take it’s share, specialize in specific sectors. Google will leave the high end corporate market to others too. Agentic AI too, maybe just keep a finger in the infrastructure and protocols. But let’s not kid ourselves. When you look at the big picture, Google is years ahead of everyone. Open AI opened Pandora’s box sooner because that was the only play they had.

  • Will AI destroy Meta?

    Will AI destroy Meta?

    I’m not even starting on the joke of the Metaverse and the billions wasted by Meta trying to convince us that they aren’t Facebook and evil personal information thieves. We don’t use Facebook to share pictures in my clan.  Long time now.  We use Google photos.  For any trip there is a shared album.  If someone says “how was it?” I share the link to it.  Nobody else sees anything.  Google can use them all they like, Google actually provides useful stuff in return. And this should be worrying Meta a lot because more and more of us are finding other ways to share on our own terms.

    So do you use AI for marketing?  Superb! AI agents will soon take over preparing social media posts.  Maybe even other AI agents exploring social media, bringing back information, adjusting campaigns.  Fantastic! Will they be as good as a good human? No, not at first. But most companies don’t use “good humans” anyway, most corporate posts are mediocre at best.

    So the problem is that social media isn’t ready for this change.

    Already the Facebook feed is almost useless.  Between sponsored posts, ads and posts of people you don’t know that are probably also sponsored, there isn’t much of interest left.  Now imagine how much worse this will get with AI agents flooding the feed.  One way to try and handle it is like my kids do.  They are extremely careful about connecting to anyone.  Person or brand.  So a curated contact list in an application that only allows for content from your contacts.  That’s a good start. But it doesn’t leave much room for Meta to sell anything. And it is more than likely that they won’t do it on a Meta platform pretty soon.

    It’s also the death knell for mediocre social media posts from companies promoting something.  And that will include AI generated ones.

    Good content is still king.  Just in a different way. AI models are training on freely available information on the web.  So get your website up to speed.  Make sure it projects your expertise and goes in depth into your products or services.  In a convoluted way it will still eventually bring customers to your door.  But only if it is quality content with meaningful individuality.  Anything else will be competing with the bot armies in a sea of crappy mediocrity, AI generated or not.

    Companies are busy trying to get their vast swaths of data into a form that can be accessed by AI in order to milk some future advantage.  So they can discover the secret winning patterns nobody noticed all these years.  Well, while you are doing that, maybe also try and make sure you have enough original content which accurately describes what value you are adding where.  Because every other AI is trying to figure you out in order to bring the “best” result to their users.

    Those rushing to pronounce that “search is dead” haven’t really thought it through. Sure, we will run AI powered searches. And how do they learn? They LargeLanguageModel themselves silly, hoovering up data and trying to make sense of it. So if I ask “which is the best lightweight sleeping bag for hiking trips” they are scouring forums and websites and anything they can find. Well, they can’t find social media posts because Meta hides those. And when it comes to making sense and prioritising all this information the game hasn’t changed. A good product which has gained fans that talk about it a lot will still win. No matter who powers your search. It’s a long term game like it always was. And Zuckerberg is at heart a low level hacker in a hurry to make a quick buck like he always was. No innovation. No essential services for users. No trust in how he will milk our data. No real structure to all the information users have thrown on to his platforms all these years. No added value.

    AI is not Meta’s friend right now.  Because at the end of the day, really, what has Facebook ever done for us?

    (My personal Facebook account got wrongly suspended a year ago. Meta never even responded to multiple requests about it.)

  • Apple can’t buy it’s way into AI

    Apple can’t buy it’s way into AI

    Have you heard of Google Crowdsource? It’s been around for years,

    a lot of volunteers helping Google become even better at AI. Thousands of people doing image label verification, or audio validation, handwriting verification and millions of other clever ways to help Google improve. I am Level35 because I found it interesting from the start. I enjoyed the tech talks from Google AI people and the seminars. You know, back when AI wasn’t all the rage like it is now.

    Or maybe you have heard of Google Photos. A ridiculous in scale endeavour whereby Google gave the entire planet (after all 8 out of 10 smartphones run on Android, not iOS) infinite photo backup. Thus sucking up the greatest free database of images anywhere. Ever. While iPhone users all have phones out of storage because they don’t want to pay anymore than they already pay every month to Apple.

    These are just two examples that you might not have thought of. Of course Google Maps, YouTube or Google Search are also on a scale way beyond anything Apple can even imagine. This isn’t like Siri, Apple can’t just throw 200 million to buy an SRI equivelant to catch up. In fact Siri is an excellent example because even when it was launched it was many years behind Google. Much like with Apple Maps, Apple seems content to have an inferior product simply window dressed for the North American market.

    Google has invested long term in making Google Maps a globally relevant product. There has been a vibrant group of contributors adding to the quality of the data and the AI. Apple has been too busy trying to upsell devices at a greater profit margin to bother to improve its products. Some say it’s the “dark horse” playing “the long game”. They are ignorant and almost funny in their ignorance.

    It isn’t just about massive datasets. AI focus has been at the core of Google’s existence from the get go. Converting Google search to Chat GPT like functionality was just the flip of a switch for Alphabet. They didn’t need to buy any companies because they are that company. And if it is about computing power again Google has the best model. They have been selling “free” services like search, maps, youtube and everything else for years, building massive infrastructure on the back of their ad business in ways that consumers consider a win win.

    Apple’s endeavours in digital video creation are an excellent illustration of the extent of the hole they have dug themselves. A few decades ago the company thought that digital video creation would be the next DTP, the next saviour for their computing division. Even now with their touted advantage in processors they have just 20% of the global computer share. Back then they bought software like Final Cut and other more expensive digital creation tools and tried to package them in various ways to sell more Macs into studios and broadcasters. It didn’t work of course. Apple was not that kind of company, it made money from iPhones. There was no durability to the vision and the software started dying out with less and less updates. It went back to being a cute toy for some Apple loving film folk.

    Right now Apple has no AI vision. It can’t even think of how to develop one. They are too busy trying to think of some gadget that will sell as the iPhone’s market share which is puny. For all you Americans reading, globally just 15% of humans use an iPhone. They all prefer products that are more affordable, more flexible, more open and with the innovations that Apple copies for their next flagship every year. They haven’t been beating world champions at the game of Go, nor folding proteins.

    So when Apple inevitably announces what its plan are in AI, I would take them with a massive pinch of salt. Buying any company won’t solve their problem and no new technology they can announce will either. AI thinking is a long term project and Apple has been absent from the party.

  • World Health solved (it is still noon)

    World Health solved (it is still noon)

    Nope. Not yet. I will not buy a smartwatch. Not smart enough. Sure, if you like wearing a watch anyway, there are plenty choices that also measure things. It’s not that I am waiting for more sensors. Even with what is available today on a $30 smart band we should be getting more useful information.

    My phone knows how much ground I am covering when walking my kids to school. The watch knows my heart rate. So why isn’t it telling me if I am getting more or less fit? My phone knows how much screen time I am getting late at night. My watch knows how well I slept. So why isn’t it telling me exactly what to change in my schedule? Hey Google, should I watch a movie or mess about on Facebook right now?

    These examples are simply scratching the surface. As I fire up Google Docs and start writing, it could even give me ideas like “go drink something, you are struggling to get a decent sentence” or “I see you can’t find a catchy title. Maybe tomorrow you go for a walk first, statistically, you come up with titles much faster on days that you walked in the morning before.”

    The only rational reason a company like Google isn’t doing this is so that we don’t freak out. It is far from inconceivable that the company knows when we go to the toilet and how long we spend there. Without demanding any user input (that’s how I like it best) we could be getting very very useful advice on diet and lifestyle that really make a massive difference. Our phone knows which doctors we visit and when anyway. Our Google searches, the apps we use, the speed with which we click or even how often we idly play with our phone’s screen unlock, all these data points, when connected, surely give a powerful insight into our health, mood and potential at any point in time. Over time and with millions of people on this platform, it is safe to assume major new roads of inquiry would open in terms of global health. If a pandemic struck the world, we would not be blind begging for more test data; we would already have it as some combination of heart rate/temperature/activity/blood oxygen would surely fit a pattern which would fairly accurately predict if you have SARS or something like that.

    The more you think about it like that, it is ludicrous that doctors aren’t demanding this data in order to make better decisions.

  • Google will never buy Kevin Durant

    When the Golden State Warriors got Kevin Durant we were impressed.  What was already a super star team just got even …super starier.  The Warriors couldn’t even hide the fact as he single handedly won matches and LeBron tried to defend 2 or even 3 players on his own at times.  As the finals ended today, (4-1 exactly as I had publicly predicted by the way) I find it hard not to draw parallels with Google.  

     

    Between Google and Facebook, online advertising has demolished the old names in media.  Hundreds of million of dollars are flowing every day in a new direction.  But only Google has the intelligence to handle it well.  It isn’t wasting time in messaging apps because messaging apps do not contain much useful information from which to figure out what ads to show you.  Facebook has the 4 most resource hungry apps for Android.  Messenger, Instagram, WhatsApp and of course the Facebook app itself.  For no good reason.  As they integrate Instagram with Facebook more and more, you can’t help but wonder why we even need four separate apps from Facebook.  While they experiment with screwing up your timelines, Google is getting the job done.

     

    Google keeps improving Android at such an amazing pace, even now, that Apple only has to copy a few features every time to manage to entertain its audience.  Which is sort of the point.  Nobody was anywhere near Google in Artificial Intelligence.  Yet Google didn’t launch anything magic until it needed to.    It could blow everyone out of social tomorrow.  Some of us remember that day when our Google searches showed us what our friends were doing and what they thought of our Google searches.  It was a freak day, maybe I was on some A/B test.  I saw what Google could do if they used more of what they know about me and my friends.

     

    Google will not though.  It is not about having 4 All NBA players on your team.  It is about hiding their talent unless it becomes necessary.  If Golden State blew out the Cavs today by 160-80 everyone would be furious.  Call it sportsmanship if you like, I call it clever business sense.  Basic human understanding.  Nobody likes a freak.  Society does not tolerate outliers.  Business and government tend to target whoever does too well.  Google will never make a fuss about buying a Kevin Durant.  They got Ray Kurzweil and didn’t make a fuss.  He got natural language recognition working way better than everyone else.  Siri is a joke but Google is making sure you don’t feel bad about it.  In fact it is quite likely that Apple and Facebook end up borrowing technology from Google in order not to fail; they are now officially too big to fail.

     

    It takes a little bit more to make a Champion.  And more after that if you don’t want everyone to hate the Champion.

  • English uber alles: the language digital divide

    Google assistant is fantastic.  Unless you don’t speak English.  In which case it is almost useless.  The whole “Artificial Intelligence” vogue is rather misleading.  Because when I speak to Google Allo I am still using all my experience in computing.  It works great for me because I think like a computer.  I break down my questions into chunks the way I think the computer wants to hear it.  I add qualifiers, words to help the machine understand with more accuracy.  I use terms that are more likely to work.  When we say “natural language” hey, there are classifications.  I use “natural language more likely to be understood by Google”.  It drives others crazy.  They blame my perfect accent.  “But I said the same thing!  Why doesn’t it work for me?”  

     

    Here’s the problem.  Google and pretty much everyone else in Silicon Valley, they are all only thinking in English.  Your Amazon Echo is designed for native English speakers.  (Pun intended.)  All your gadgets are.  Worse still, the intelligence is designed around people thinking  in English.  All the structure, the concepts, the way it is set up.  It is rather entertaining how some people get caught up with the fact that slang and tech words are conquering the world.  That is the tip of the iceberg.

     

    Silicon Valley is moving ahead of the rest of the planet with leaps and bounds.  Light years ahead.  We don’t have local information.  We can’t use amazon like you do.   We can’t pay for stuff or call a self driving car.  Amazon will not be able to deliver to the trunk of my car either.  The United States are a test bed for new tech and the gap with everyone else will grow exponentially.  And only in 2030, when computers are smarter than humans, maybe, just maybe, those computers may decide to develop all these wonderful tools for the rest of the earthlings.    And even then it will take a lot of work.  Because English is the language that provides the structure and concepts.  More likely that you will have all learnt to think like Google by then. 

     

    2030 is still material for science fiction.  Today, now, it is clear that we all have to move to the Valley or fall behind.  We have neither the data with which to develop such advanced tools, nor the number crunching power.   The entire planet sends their thoughts to Google every day  Out position, habits and preferences.  It is no conspiracy theory, it is simple mathematics.  Not impossible to catch up, just really really hard.

  • Stealth – when Top Gun met A.I.

    There have been many movies about Artificial Intelligence lately.  Most of them pretty good actually, generally quite thought provoking.  Especially for those of us who regularly talk to our phones and rely on cloud services for daily stuff.  Close to home as they say.  But none of them were fun.

    P-40:
    In the good old days, if you wanted character in a plane, you painted a mouth on it!

    So here are three hitech warplanes who get a new team-mate.  An all singing, all dancing, all watching and listening plane that learns on its own.  If you know very little about drones, you might think it is somehow close to reality.  If you know anything about military protocol you will not.  But that isn’t the point.

    The First World War was the first time air warfare had played a role in combat and this picture of French warplane, Caudron G3, was captured by a photographer in 1914: Everyone says “Top Gun” when you speak about plane movies, but by the end, this is more like “ET” meeting the “Iron Eagle” trilogy.    Epic stuff.   People smile at the right time, salute, explode and laugh just when they should be.  Well made movie.  For its kind.  As long as you don’t try and relate anything you see to anything in the real world, geography, politics or technology.   They even threw some romance in for good measure.

    This is not for everyone.  If I wasn’t in the right mood, it wouldn’t even be for me.