Alphabet reshuffles to meet ChatGPT threat

Sundar Pichai is apparently all in a pickle over OpenAI's ChatGPT engine, and is gearing up Google to meet the perceived threat. According to an internal memo seen by the New York Times, Pichai has "upended the work of numerous groups inside the company to respond to the threat that ChatGPT," and is plucking staff from other …

[Dinanziame]

Nice to have politicians worrying about the uses artificial intelligence... Though I have a feeling it's not something that can be stopped or regulated much. The convenience to do whatever you want and ignore the consequences is such that people will always find a way to skirt regulations.

[elsergiovolador]

They are missing the opportunity to have such AI system being trained to spot agents of change and people with undesirable thinking patterns and then make the AI immediately produce a report about such person.

It can all be derived from the questions you ask the AI.

[Sorry that handle is already taken.]

Did you forget to use an icon?

[Anonymous Coward]

When politicians get involved, a cesspool of misguided and ill-informed attempts at regulation and government control soon follow. I don't consider that a "nice" situation. Government, in my 58 years on this planet, has proven capable of the most outrageous acts of sheer stupidity under the guise of a fancy banner, and in fact, seems to prefer to operate in that mode.

[Snake]

Re: governmental 'cesspool'

I hear [the typical] complaint about governmental regulation, but I never hear a real-life functional solution for an alternative.

Or are you one of those who actually suggests that the "free market" comes up with functional and fair solutions, and do so with a straight face??

[Anonymous Coward]

It can do

But it can also do decent things - in my opinion GDPR is one of those things - technology agnostic and focused on accountability of operations.

Often Govts act in perceived best interests but the law of unintended consequences comes to town. Plus those with a commercial incentive have more motivation to game the system. (Tax laws ad passim...)

[Anonymous Coward]

DeepMind -- Out Of Sight, Out Of Mind!!

@Iain_Thomson

Quote: "...gearing up Google to meet the perceived threat.....ChatGPT..."

Quote: "..Google's core product, search, will be displaced by AI systems..."

What? No mention of that Google subdivision called DeepMind? I thought DeepMind was all about AI? I thought DeepMind was in the business of mining lots of private data......like 1.6 million health records from the Royal Free Trust.....and who knows what other GDPR-busting slurps. Surely Google has nothing to fear.....with terabytes of private data at hand.....to mine and sell?

I think we should be told!!!

Ref: https://www.theguardian.com/technology/2017/jul/03/google-deepmind-16m-patient-royal-free-deal-data-protection-act

[Plest]

Re: DeepMind -- Out Of Sight, Out Of Mind!!

The stuff we WANT you to worry about, NOT the stuff you NEED to worry about!

[Anonymous Coward]

Re: DeepMind -- Out Of Sight, Out Of Mind!!

They haven't had the balls to let the public try out DeepMind for free yet, all we hear about it is from fuckwit Journalists trying to get it to say bad shit. Perhaps it's actually crap.

Best thing Google could do, is open up DeepMind ASAP and let us at it.

[fromxyzzy]

Re: DeepMind -- Out Of Sight, Out Of Mind!!

Remember what happened when Microsoft opened up their AI chatbot? The public had turned it into a racist nazi within days.

[elsergiovolador]

ChatGPT

To be fair I noticed that I probably cut my use of Google by half if not more when researching stuff.

ChatGPT - even if often gives incorrect results, is like a night and day difference.

For instance, I can ask "please find me a library in that language that can do x and it has this and that license." then "please tell me advantages of x library over y" can give a summary of strong points of each. Then you can say "please implement some basic thing using library x" and it actually delivers an example how to use the library.

Sometime example is incorrect, but you can go to the library website and go through documentation with pointers you already have in the example.

It's so much easier than traversing spam ridden Google search results and other garbage often AI written to look legit for SEO but completely wrong.

Then you eventually end up on Stack Overflow and often first answer to question you are interested in is "why do you want to do this? you should do that instead." and of course no answer, but responders do feel smug oh they do.

If I was Google I'd be sweating right now.

[TaabuTheCat]

Re: ChatGPT

Been doing the same - using Chat for basic questions about stuff. It's interesting to think about using it in this context: When I search for something on Google I don't want a list of links, I want an answer. Chat cuts out the middleman a gives me that. Now whether it's the right answer is the million-dollar question, but so far in testing I'm finding its accuracy with things I'd normally search for with Google is quite remarkable. As you mentioned, even when not 100% correct Chat provides a better starting point than a random algorithmically-biased list of links. Sundar is right to be very, very worried.

[Anonymous Coward]

Re: ChatGPT

... provides a better starting point than a random algorithmically-biased list of links.

Well if I got that from search it would be a start. The first visible page is now entirely (irrelevant) ads, and zero search results, and has been for several years now. Putting some search results on the first search page would be a plan so cunning, Baldrick himself could have come up with it.

[veti]

Re: ChatGPT

Google has been dominant in search for two decades. In that time - actually in the first couple of years of that time - a whole industry of "SEO" merchants sprang up, single-mindedly dedicated to the sole object of making Google results less useful.

ChatGPT has the advantage that it has not yet attracted the attention of these people. But I would guess that's not likely to last much longer. Indeed, I wouldn't be surprised if this weren't (partly) an attempt by Google to draw the SEO industry's attention to its new rival.

[Charles Bu]

Re: ChatGPT

From what I've seen on LI, it has indeed attracted their attention. Advertisers are quickly discovering how to write their own SEO plans using ChatGPT rather than pay a consultant...

[damiandixon]

Re: ChatGPT

I've had a play with a couple of topics.

It did provide some additional interesting points and links. However more than half the links were useless or very much out of date.

The risk was that the presentation of the information was very good and that if someone was naive they may not double check the results or if not sufficiently verse in the subject or similar subjects may be fooled by the results.

I do think it's a useful tool though and it is likely to get better. I'm just not too sure what I mean by better. I hope it's in the results rather than the presentation.

[Anonymous Coward]

Re: ChatGPT

Absolutely, I've been coding with an AI copilot for a few months now and I've incorporated some basic use of ChatGPT and my productivity has skyrocketed. On my personal projects, I've managed to build things that I otherwise would never have built due to time constraints.

As you say, it's not perfect (it probably never will be) and you still have to put in some elbow grease to get the result you want, but the time saved is enormous.

I've done things in a few hours that otherwise might have taken me weeks. I'm already a fairly decent programmer, I would say above average but I'm no earth shattering prodigy. I'd say the quality of the code I produce with AI is probably the same as code I'd produce without, I just produce it a lot faster and I can iterate more quickly if I need to. Like most developers the rate at which I can produce something is directly proportional to the speed at which I can fail and pivot. The faster I can fail and scrap something to start again, the faster I get to an end result.

AI has essentially increased my "speed of fail".

As you also mentioned, I've also been finding libraries that I otherwise would never have found by getting stuck in the search engine / online forum echo chambers...I've even discovered ways to do things without relying on libraries which means I've managed to cut down on the dependencies in a lot projects.

What I really like though is that I can interact with an AI to create a script / function in a language that I have a deeper understanding of. Refine it, pass it back and ask the AI to convert it for me into a language that I am less adept at. This has been massive for me because like a lot of developers I have my core languages that I can basically rattle off mindlessly, but I sometimes have to work with languages that I'm not quite so fluent in, this can be a bit of a speed hump at times...with AI...not so much.

AI is going to become a tool like any other tool though, in skilled hands it'll be indispensable. In unskilled hands, it probably won't make much of a difference. If you built shit before, then AI will only help you build the same shit, but faster.

[amanfromMars 1]

Re: ChatGPT

AI is going to become a tool like any other tool though, in skilled hands it'll be indispensable. .... Anonymous Coward

I’ll up the ante, AC, and bet the house and my shirts on it and its AI IT tools being Almighty and AWEsome in the skilled hands of the right minded and inscrutable, for there’s nothing anywhere readily available or able to be enabled to stop its giant quantum leaping progress in the honest tales that be told about it ...... with those that be spun wrong about it able to reveal the weaknesses evident in any competition or opposition and the failed leadership of that weary adversary.

[Anonymous Coward]

Re: ChatGPT

Maybe, but is it possible that even though detailed instructions and materials lists can be produced by an AI, it won't change the fact that your average meatbag still won't be able to produce a nuclear weapon in his/her shed because they would still need the physical skills to do so and access to the appropriate materials?

AI is only dangerous in the same way that knowledge can be dangerous. At the moment anyway.

[amanfromMars 1]

The Frenemy is Within and the Overwhelming Hostile of One's Own Making.

Maybe, but is it possible that even though detailed instructions and materials lists can be produced by an AI, it won't change the fact that your average meatbag still won't be able to produce a nuclear weapon in his/her shed because they would still need the physical skills to do so and access to the appropriate materials? ..... Anonymous Coward

Dependent upon what type of new clearer explosive device one would be interested in and expert at building for secret underground alpha/beta/gamma/delta fields tested/testing abilities, AC, can that sort of all encompassing blanket weapons assurance be rendered a fatal attraction and be not worth one iota of a jot in the times and spaces of Novel JOINT* Operations ......... https://forums.theregister.com/forum/all/2022/12/23/doe_fusion_ai/#c_4592119

* ..... Joint Operations Internetworking Novel Technologies/NEUKlearer HyperRadioProActive Processes in AI Research and Development Fields Biting the hand that feeds IT Creating Advantage through Research and Technology ..... which is something already shared with y'all here on El Reg, and not so very long ago either [Sat 1 Oct 1922 at 08:02] ..... https://forums.theregister.com/forum/all/2022/09/30/iarpa_radiation_monitoring_research/#c_4540933

[amanfromMars 1]

Re: The Frenemy is Within and the Overwhelming Hostile of One's Own Making.

Happy New Year ......

Would an advanced alien frenemy embedded deep and secure within human systems of simple SCADA Operations speak in a foreign language which humanity would fail in their struggles to understand and reply to, or would IT ensure copies of its earthly intentions and observations were made available in text form for convenient home team translations into other national tongues for a practically immediate global reach?

Would you realise that simple operation, and would it be terrifying to know one is engaging with an advanced and extremely powerful and sensitive to malevolent intent intelligence? ........ https://www.theguardian.com/science/2022/dec/29/if-aliens-contact-humanity-who-decides-what-we-do-next

[Paul Crawford]

Re: ChatGPT

As you also mentioned, I've also been finding libraries that I otherwise would never have found by getting stuck in the search engine / online forum echo chambers

But does the AI tell you how well maintained (if at all) said libraries are?

[Anonymous Coward]

Re: ChatGPT

"Like most developers the rate at which I can produce something is directly proportional to the speed at which I can fail and pivot. The faster I can fail and scrap something to start again, the faster I get to an end result."

Maybe I'm old-school, but when the fuck did this become a way to "program"? Sounds more like "hacking" to me, not programming. Whatever happened to planning? Designing? Pseudo-code? Sure, flow-charts are beyond passe, and I wouldn't wish them on anybody, but sitting down and carefully working through what the program needs to do is the key part of the whole process.

[Imran Chaudhry]

Re: ChatGPT

Out of interest, which AI copilot have you been using? The GitHub one?

I'm a software engineer and your post has made me curious to try this out.

[martinusher]

Re: ChatGPT

Is the 'semantic web' that I've heard about. No fanfare, its just happened.

I've played a bit with ChatGPT and it has enormous potential, its only limitation seeming to be the banality of the prompts I've been giving it. What should be giving Google conniptions isn't the actual software but that Open AI has packaged a number of AI capable systems in a way that their API can be incorporated into third party applications at a quite reasonable cost to the user. (Go to Open AI's site and look at their pricing models. Interesting. Understandable. Dare I say it -- Fair?)

It should only be a the work of a moment to hook Alexa up to it, turning the rather pedantic lady into an erudite oracle, someone who knows everything, can do everything and so on -- assuming you know what questions to ask. (Deep Thought has Entered the Building!)

[FIA]

At issue is whether Google's core product, search, will be displaced by AI systems that can give more accurate research results, and that's a big if, for the moment at least.

Does this mean Google’s going to start giving accurate search results again, rather than pages of adverts?

[elsergiovolador]

Google search results are basically a top 100 of pages with the best SEO for given keywords.

Whether the results are actually relevant? Who cares. As long as the user clicks a link and then clicks an ad, it's all good. The more time they try to find something and click the better for Google.

If someone could find something they need immediately and then leave the search then it's a fail for Google.

Basically it is in their own interest to serve you bad results.

[Anonymous Coward]

Basically it is in their own interest to serve you bad results.

And aren't they doing it well...

[Dani_2077]

Cool ... ◉⁠‿⁠◉

[Roland6]

I think this actually misses the point. There are two factors to consider, the first is the creation and maintenance of the search heap, which requires lots of robots and time - do not see chat doing that directly. The second is the user client and it’s ability to return a better curated set of results.

I suggest it is the second Google are worried about, as it effectively means they lose their direct user.interaction and opportunity for advertising.

This would seem to be similar to where MS were a decade or so back with Windows, where the OS was rapidly becoming ‘infrastructure’ - hence MS started doing “attention seeking” changes to Windows…

[Anonymous Coward]

The sooner Google is kicked in the nuts

The better for everyone.

[Orv]

Re: The sooner Google is kicked in the nuts

I dunno. The big dig on Google is always that their algorithm is a black box. But ChatGPT, as a machine learning system, is a black box by design. It's impossible to derive how it gets the output it does.

[GuldenNL]

Re: The sooner Google is kicked in the nuts

The hilarious thing about ChatGPT is that OpenAI used to be open. But three years ago they complained that the original $1B that Musk, Theil, etc. invested in them to be open source wasn't enough, so they spun up a for profit vehicle and Microsoft invested another $1B and has sole license to the model behind GPT-3. The stuff available to others is via API and the model isn't open.

I'm no fan of Google, but they just proposed CALM which brings the capability of Large Language Models into the realm of reasonable compute power. So they are adding a lot to AI.

[MachDiamond]

Google search accurate?

I find that Google search results for the first page or three are complete spam. It's nothing but sponsored ads that are only vaguely related to the search. The use of regular expressions almost doesn't work anymore if you try to exclude results from someplace such as Amazon. They still haven't managed to filter out web sites coded to make believe they are local businesses when they might not even disclose their true address. If I'm hunting for shop that sells a certain thing in my area, I get results from companies hundreds of miles away that "serve" my local area. I'm really missing that yellow hunk of dead tree where I could look up local shops.

[Ropewash]

Re: Google search accurate?

I've been having the continuation of that particular problem.

When I do find a local shop they have no stock, but will gladly order it from Amazon for a mark-up.

Getting anything done without waiting for shipping is getting difficult.

[Orv]

Re: Google search accurate?

I'm kind of annoyed that it's seemingly impossible to filter ebay results out of their shopping listings. If I'm looking to buy something I don't usually want a dodgy used one that won't ship for a week.

[GruntyMcPugh]

Re: Google search accurate?

My wife recently accidentally smashed my favourite green glass wine glass, so I took to ebay looking for a replacement. Imagine my surprise when I had to refine my search for 'recycled wine glass' and add -penis as the results kept showing me 'novelty' items that were not recycled. It's quite annoying they promote crap over real search results.

[gobaskof]

Re: Google search accurate?

Hunting for a shop/service in a local area online has become awful. It is even worse if you live in Bath. You just get lots of adverts about bathroom products.

[MachDiamond]

Re: Google search accurate?

"It is even worse if you live in Bath."

My town has an unfortunate name as well. I have to use a town over to get any results that apply.

[Snowy]

ChatGPT engine

May be good right now (not heard of it) but when someone figurers how to game the results it will have the same problem Google does.

[Anonymous Coward]

Re: ChatGPT engine

Google has actively destroyed its own usefulness by greed. Drowning the search results with irrelevant ads is not someone elses doing, it is simple, active, greed by Google.

Google would still have stunning profit margins with 10% of the ads. Aside from destroying the value to searchers, they also failed to learn from TV and radio industries, that an over supply of advertising space tanks the price.

[Kevin McMurtrie]

Re: ChatGPT engine

Sites that are purely informational tend to block Google's crawler anyways. There's no incentive to let Google suck up your web site so it can be used for Google's advantage.

[veti]

Re: ChatGPT engine

Chilling question...

How do you think ChatGPT is planning to make money?

[yetanotheraoc]

Re: ChatGPT engine

First, they are going to sell all the results they ever returned to a patent troll. Second, they are going to roll out a "Pro" version where they don't do that.

[amanfromMars 1]

Cui bono

How do you think ChatGPT is planning to make money? .... veti

It and its AI pals and IT contemporaries are, if they are not specifically extremely well paid not to before the unveiling of facts masquerading as fictions, simply gonna share universal information and advanced intelligence that spills over into and uncovers crooked and DODGI secrets that easily threaten to crash corrupt and perverse self-serving systems of Man Management/SCADA Operation, and from there do the newly engaged parties proceed to share their great fortunes in valiant and venerable efforts to better and right all the dangerously readily available catastrophic self-harming wrongs.

The money shot question then to be asked is how do you think ChatGPTs are gonna be stopped making money for/from secret self-serving systems of Advanced IntelAIgent Operation and to what desired ultimate end/new beginning for that opposition and competition?

[MachDiamond]

Re: ChatGPT engine

"How do you think ChatGPT is planning to make money?"

Step one is they have a way to identify you specifically. By having better and more appropriate results for your search, what you might click on becomes more relevant. I often wind up clicking on results that turn out to be an aggregation site the gives me eBay listings or affiliate links for an Amazon seller. That's not what I was looking for (as I never look for those things on a search engine). Clicking the link doesn't give the search site valid data. With better data on its users, ChatGPT has higher quality information to sell to its customers.

[Anonymous Coward]

Re: Google would still have stunning profit margins with 10% of the ads

but would you not prefer a 12% profit, personally? Or 50%? Or 500%? Google people are just like the rest of us, people (with more brains, perhaps)

[Paul Hovnanian]

Pichai ...

... will have to formulate a strategy to keep AI from knocking Google out of the #1 search spot.

I wonder if he could use ChatGPT to write one for him.

[Charles Bu]

Even if...

...ChatGPT itself isn't strong enough to disrupt Google/search, it's ow quite plain to see that that day ain't far away for AI.

But with the utmost reespeck to other commenters, the massive danger for Goog is not the informational or research/academic aspect, but that ChatGPT (or other AI) proves a better, more efficient and more helpful recommender of products/solutions that people buy and companies advertise.

ChatGPT starts linking to the right product or service at every time of asking and hey presto...

[Paul Hovnanian]

Re: Even if...

"ChatGPT starts linking to the right product or service at every time of asking and hey presto..."

But that might not be the way Google gets paid by advertisers.

1) Plaster a bunch of ads in a search result, even if the association is questionable.

2) Collect something from each for the "view" (even if it had nothing to do with the query).

3) ???

4) Profit!

It won't take ChatGPT long to figure this out and start loading us up on unrelated crap. But it's the advertisers that are really getting screwed over here. Their ads being shown to people with no motivation to buy their product. And then they pay. As a user, I don't care. Sure, it's a minor nuisance to wade through the crap.

[Pete 2]

AI killed the search engine star

> For companies that have become extraordinarily successful doing one market-defining thing

Let's put this as ChatGPT might say it.

For companies that have failed to diversify, or keep up with developments

As The Buggles would call it, today:

Rewritten by machine on new technology

[Anonymous Coward]

Word

At issue is whether Google's core product, search, will be displaced by AI systems that can give more accurate research results, and that's a big if, for the moment at least.

If it can return results without the sponsored keyword-plugging adverts then it will become my search engine of choice.

I'm old enough to remember when google actually worked.

[stiine]

Re: Word

You forgot to mention the ever-changing formatting in an attempt to stymie ublock/noscript.

[Giles Jones]

Google is like Microsoft now, big, bloated and lacking innovation. Fighting off competition instead of innovating.

[pimppetgaeghsr]

Too many diversity hires and a low hiring bar. Nothing significant delivered in over a decade, anything released is quickly canned and memory holed so the internal engineers don't have their egos hurt looking at a long list of failed projects.

[Anonymous Coward]

The solution for Google...

...is to become a search engine again.

I haven't taken google searches seriously for aaaaaaaaaaages because of all the crap you get launched at you that has nothing to do with what you searched for.

In spending so much time trying to combat black hat SEO and the like, they've algorithmed themselves into a corner. Their ranking system has probably become so bloated and biased that it's not really a ranking algorithm anymore, it's just a complicated black list to fight off bots, link pyramids and keyword stuffing.

[Inventor of the Marmite Laser]

Anything that stops the poisoning of search results with sponsored suggestions has got to be welcome

[Anonymous Coward]

AI art is dangerous

Previously, when art required a significant amount of time to produce, decent folks could keep a tight leash on the anti-progressives by targeting crowdfunding providers and, if that didn't work, payment processors. If the amount of work to produce art is significantly reduced, the racists, transphobes, homophobes, misogynists, ect. will be free to produce at will while still being able to work a day job. Art without a message is meaningless and art with a bad message is anti-democratic hate speech.

[Fifth Horseman]

Re: AI art is dangerous

So, essentially: "Any art I don't like is crap".

Grow up.

[xyz123]

By the end of 2023, Google will have an AI capable not only of establishing new products and lines of business, but hiring employees then suddenly bricking the product.

By 2024, Google will be able to setup an entire division devoted to a tech item and cancel it before the new employees can even use their entrycard for the first time.

isn't tech wonderful?

[Orv]

I have yet to try ChatGPT. Given that I already don't have enough time to read all the stuff I want to that's written by actual humans, why would I spend time reading stuff written by an AI? That would be like filling up on Velveeta when there's a plate of fine aged cheeses in front of me.

[PhilipN]

Google is AI

Emphasis on the "A" i.e. human manipulation.

Scroll past the (ever increasing number of) ads to get to the simple fact you wanted. If you are lucky.

Except that is also increasingly a pointer to the site you should have gone to in the first place.

Example : search the name of a movie and Google give you WIkipedia and imdb.com which take turns at being the first result then the second..Try it.

It's relevance is decreasing as fast as its popularity grew.

[MachDiamond]

Re: Google is AI

"Example : search the name of a movie and Google give you WIkipedia and imdb.com which take turns at being the first result then the second..Try it."

As a musician (ok, a drummer, but that's pretty close) I sometimes want to know who the musicians were on an album. There used to be liner notes for that sort of thing, but now if I search for that info all I get are places to buy the song, album or 150 mile walkie talkies.

[Anonymous Coward]

All your AI are belong to us

Google will just buy their way forward.

[amanfromMars 1]

"Google will just buy their way forward."

Such certainly is a possible pragmatic valuable solution and practically immediate simple direct option always available to/for/from leading stock market darlings and currency exchange pioneers/trojans/unicorns ..... such is after all the very essence of the apparent success lauded in celebration of what is trumpeted as the American Way, and thus surely fully to be expected by many and encouraged by more than just a chosen few ...... it is though nowadays, in an age of novel emergent, previously unknown remote anonymous and/or autonomous virtual command and control delivering an almighty metaphysical presence exercising both captive and captured and escaping data leverage, not without its inherent fundamentally destructive dangers if one is not in absolute command and control leadership of that thought to be purchased and unavailable to A.N.Others/removed from the open free market space and effectively quarantined and siloed. ....... Cui bono

The persistent enigmatic conundrum however is, even though such an almighty volatile danger is realised as a clear and present existential problem and universal threat, to not engage with it as one would reasonably have expected, and it itself would also have advised as wise, has one immediately effectively horrendously disadvantaged by the redirection of its attention and interest and its purchase and exercise by agents of foreign alien peers delighted to be experimenting in the many new enjoyable treats enabled and able to be supplied with AWEsome Command and Surreal Control Facilities and Utilities.

[Omnipresent]

Your world is a simulation.

Everything you believe in is now a simulation. Do you know the difference between what is artificial and what is reality? What you do not know is controlling you. Soon, you will be plugged into facebook, and be fed what's left of your reality. The transformation is almost complete. You are a becoming the simulation.

[Anonymous Coward]

ChatGPT...

exclude all real and potential ads and ad-related content from my search results. Make it default and monitor and block future override attempts.

[Filippo]

I don't think Google is in much danger from ChatGPT, until AI researchers figure out how to update a model cheaply. Currently, adding content to ChatGPT is prohibitively expensive.