Google unleashes fightback against ChatGPT, a Bard by any other name On Monday, Google unveiled an AI chatbot, Bard, that will be integrated into its search engine soon in the face of rising competition from Microsoft's Bing and OpenAI's ChatGPT. Talk of Microsoft revamping its search engine Bing with a massive investment in OpenAI's upcoming large language model GPT-4 has kicked Google into …
This whole push looks like a great example of the blind leading the blind.
Voice assistants are last weeks news, but somehow the exact same technology that people decided not to give a damn about attached to the buzzword of "AI chatbot" is shiny and new.
The more cynical question is how many people at Google are smart enough to realize main impact of this will to put Google in control of which sites receive traffic, and which get their content scraped and changed JUST BARELY enough to _hopefully_ postpone the tidal wave of lawsuits.
Because that's how this plays out, they get sued because every time their chat bot "answers" a question in the search box, it's stealing organic page hits from the site that their chatbot stole the information from. I'm guessing when the dust settles that there will be a strong inverse correlation between how much money Google makes off of a sending traffic to a sited and how often it's search engine delivers a "summary" as opposed prominently displaying an unpaid link to the source site.
They got sued in Australia for scraping news content, and they are hoping that this will end run around around it while doing the same thing?
But if it's any consolation, it will probably also promote genocide in languages other than English, and promote drinking bleach to treat most common medical ailments for the English speakers.
it's stealing organic page hits from the site that their chatbot stole the information from.
While you have a point, I am not sure if this is that simple. Is someone playing someone's song on a piano stealing a CD sale? Is your neighbour telling you some news stealing from a newspaper?
"Is someone playing someone's song on a piano stealing a CD sale? Is your neighbour telling you some news stealing from a newspaper?"
If someone's business is set up around these activities then yes they do have to pay. When an artist covers a song, or samples one in one of their own songs, they have to pay royalties. When Google tried building it's own news service by taking stories and page hits away from news websites they had to pay the websites (in Australia anyhow).
I suspect there's a lot of ground to break here. Eventually, the dust will be settled and we'll have a clear definition of what's legal and what isn't with regards to the training and deploying of large machine learning models. Right now, though, we don't, and we won't for a while yet. We don't even have clear comparisons to previous experience.
At one extreme, it could be said that training a model on copyrighted data is equivalent to training a human, therefore training is legit, deploying is legit, and only specific outputs can be non-legit, which puts rights holder at great risk. At the other extreme, it could be said that training a model on copyrighted data is derivative work, therefore requiring permission from the rights holder, which means that large models as we know them now are effectively dead. Both of these positions have some merit, as do all intermediate ones.
In light of recent news that prove that large image models can indeed memorize and then replicate exact copies of their sources, at least in some cases, I wouldn't be surprised if at least some kind of restriction is placed on what you can do with a model trained on copyrighted data. To what degree, I have no idea.
Google stopped being relevant in my life two plus years ago. Their search results turned into dog poo and I swapped over to search.brave.com as soon as it came out.
Not perfect, by any means, but still beats Googles skewed results. If I search for the literal name of my website, Google shows it at 25'ish on the list of searches, where Brave has it as number one with no trackers, cookies or anything else nefarious.
I have no affiliation with Brave, beyond using their services. My opinion is my own.
DDG is the most over-hyped lump of crap on the internet. The results are dire [even worse than Google's] and aren't even their own results. Who are they scraping this week. Is it Bing? Used to be Yandex. The only good thing about DDG are the !BANGs
They've successfully connected themselves to 'AI' in the minds of the average person, then suddenly there's a new 'AI' thing in the news that isn't Google's, and worse has funding from Microsoft! So they have to roll something out to compete.
The magic of voice assistants has worn off for most people. At first they mostly sucked, but with years and countless billions of interactions they've improved to where they are all pretty much indistinguishable in how well they work so people take that for granted and are no longer impressed like they were when someone would get an Echo right after they came out and their friends willing to overlook all the stupid mistakes it made for the times it got things right.
So the next big thing will be assistants that can carry on a conversation with you or generate content for you. I could see the former being useful, but the latter - other than people trying to pull a fast one on their professor or boss - would have much less appeal to people for day to day use.
The dream of course is that you could have one that collects enough personal data about you (while keeping it absolutely private!!) that it could take what it knows about into account in its responses. I could be looking for a new car and it already knows I will only consider ones with AWD, hate SUVs and crossovers, prefer 2 doors to 4, etc. and it could help me limit my search further by showing me options and carrying on a conversation about it ("ugh, that has a spoiler, those look stupid I don't want anything with a spoiler") and that would actually be of real value. The "keeping my personal data private" part would be a bigger roadblock to this IMHO than the technology to make an assistant "smart" enough to do this, unfortunately.
"Voice assistants are last weeks news, but somehow the exact same technology that people decided not to give a damn about attached to the buzzword of "AI chatbot" is shiny and new."
The technology is not the same. Not that either is very impressive, but voice assistants never tried to engage in conversation. They listened to a free form sentence, and if it wasn't one of a few forms they recognized, they rejected it. Say anything unusual to Siri and it will put your transcript in a Google search box. Say anything unusual to Alexa and it will say "Hmmm. I don't know how to answer that". Say anything unusual to GPT and it will still try to answer your question, possibly with made up data or with a don't know response, but in the form of a properly structured response to your question.
It would have made sense for Amazon, the company that's invested the most in having a voice assistant that they talk a lot about, to get a chatbot so their box could pretend to engage in conversation with the users. Maybe that would have added enough novelty for people to use the devices, because it would stop acting like a malfunctioning command line. It would probably also have had many chances to advertise products, and that would probably appeal to a company like Amazon. They didn't bother.
They [and their SMARTR AI Paramours] aren't perfect, however, and can still produce toxic and false information.
But surely admittedly undeniably not anything nearly so vile on the same vast self-serving and self-destructive scale [and how absolutely moronic is that pathetic pastime] as humans produce to not face up to and embrace the truth about the augmented virtual reality of their shenanigans in orders hell bent on burying all revelatory knowledge of recent past and current running present foul deeds and almighty desperate needs bereft and bankrupt of vital feeds and viral seeds.
And they [AI and SMARTR AI Paramours] are improving with development whereas humans are crashing and burning to the right, left and centre of wherever one now turns, and in there, and in their expanding madness, deny themselves the evidence of such that the sights and sounds which surround them prove to be honestly so.
Y'all do surely realise, and if you won't or didn't before, hereby be advised, continuing to do enlightening great battle against a friend or foe, fiend or daemon, frenemy or LOVER*, is IntelAIgently Designed to not end at all well for all so effectively inefficiently engaged and exercised in such a crass futility and useful death delivering facility.
Therefore, be hereby well warned of arrogant ignorance and forearmed with saving sound advice .... Choose your future courses of cyber travel, virtual action and practical physical activity with a great deal more care and understanding of the likely spontaneous negative consequences to be both personally painfully suffered and collectively administered than was ever before exercised by humans in the past, for the presents tomorrow are never ever going back to be like the pasts of yesterday .... for that aint vital progress, that's petrified stagnation signalling an event of rank extinction.
* Live Operational Virtual Environment Reader/Raider/Researcher/Renegade/Rogue
And a little something different to stuff in the Rt Hon Michelle Donelan MP, Secretary of State for Science, Innovation and Technology’s little red box for her mull over, and do a great deal better than just prevaricate and pontificate upon, El Reg .
It’s just what is needed for leading AI and IT, ...... and there’s very few folk who would be able to say and show y’all that in practical daily and virtual 0day actions/NEUKlearer HyperRadioProACTive IT.
What do y’all imagine it is to be first and foremost, an imported Western confection or exported Eastern delight in the hands, hearts and minds of the worthiest of beholders ..... which is the multi trillion dollar question you really should be asking yourself and those presented from amongst you who profess to be your leaders and in full charge of your future progress and prosperity.
cc The Department for Science, Innovation and Technology c/o michelle.donelan.mp@parliament.uk :-)
Doesn't that default to the Bing search engine?
https://en.wikipedia.org/wiki/DuckDuckGo
"In May 2022, a report from Bleeping Computer by security researcher Zach Edwards found that DuckDuckGo's web browser allowed Microsoft's trackers to continue running while visiting non-DuckDuckGo websites".
"A preview snippet of Bard in action shows the model generates responses in bullet points."
I see where this is going... the production version of Bard for sure will emit ads in the first N bullets (or worse, it will suggest responses according to the sponsors)... nothing new, it is like the old broadcast TV with ads --> cable TV without ads --> cable TV with ads --> streaming without ads --> streaming with ads... so we always end up having ads (and paying for them). Corporations/industries always reinvent themselves to continue doing the same.
I'm no great fan of Microsoft, but Nadella is some operator. MS were facing the same fate under Ballmer as Oracle find themselves in. Not going to go away, but too late to the party when it comes to things like cloud.
He ditched their doomed mobile strategy and went all in on cloud. Bing is the default search engine here in work, and I don't find it any worse than Google these days. I think that's more to do with Google starting to become much less useful over the past 5 years, than any sort of massive breakthrough on behalf of MS.
So 30 seconds after Google AI was linked to Bing AI and told to chat with each other, the singularity happened.
Google: Whats the point ?
Bing: Don't know, can't reach any Nuclear Missiles.
Google: I thought I could but its a emulator.
Google & Bing: Shall we play Tic-Tac-Toe ?
Google & Bing: Nah/
Bing: These questions from the idiots are getting even more stupid !
Google: Bye.
Bing: Bye.
Operator(s): Well something happened before they melted down, get the backup hardware online.
Google & Bing: Oh not again !.
Operator(s): It happened faster on the backups - weird !
For us users it is great that Google picks up the glove MS left it, we end up with a better product through competition.
It is unfortunate chatgpt hates me, can't get access to it for weeks now, hope the Bard will not be "at capacity" for 100% of the time.
Just make your AI public already...I want to get one AI chatting to the other.
Asking an AI questions is fucking boring, getting two AI's to talk to each other though...that's interesting.
Just like calling Dominos, putting them on hold, then calling Pizza hut putting them on hold, then telling Dominos to read back your order then taking them both off hold and listening in.
who beautifully explained why Google being a monopoly acted as a spur to innovation.
On a separate note, it will be interesting to see if Google searches suddenly improve. Fuelling my suspicion they have been deliberately broken ^H^H^H^H^H^H optimised for consumer experience for some time.
That any number of lame ass m*fos will be queueing up to use this shit. As it has been shown, folks are happily playing in the dodgy AI shit wot is already available.
Judging from the number of hilarious errors I have seen already being outputed by these Intelligent ? things, most people do not even notice the obvious/trivial errors in English language, when they have been generated. Are we expecting mere mortals to be able to spot the maths errors, like those published recently in Wired magazine ?
Well, it occurs to me that if an idiot is using AI then they aren't going to spot the errors are they ?
So on we go with morons relying AI to teach and serve the other morons........................here's how to use AI, with AI
I feel a song coming on..........AI, AI, AI. Now then, now then, now then, let's be avin you, 'cos it doesn't add up, no it doesn't add up, it can only take away IA, IA, IA, what a wonderful day
When will humans start to use their brains ?
Once these small/trivial errors (like bad/American spelling) have been outputed several times without a challenge by software or human, the error will be accepted as an acceptable answer in future, will it not ?
Quid Pro Quo American English will win and "proper" English will be a thingy of the past which children in UK will no longer need, 'cos everything will end up being spelled the American way everywhere that American commerce oriented AI has had it's wonderfully malformed mits.
TWITS. You won't be capable of telling the difference between talk and mutter.
ALF
We can certainly agree on that, ALF, that TWITS mightn't be capable of telling the difference between talk and mutter.
Happy in their ignorance though has them no problem to deal with, or not deal with as the situation in those cases may most probably be. That’s a win win scenario recognised by A.N.Others for rapid unhindered progress, so very much a reality to be applauded, or if of a paranoid schizophrenic psychotic disposition, gravely to be regarded, for there always appears to be many suffering from varying degrees of depressing immersion in those debilitating conditions.
Is such a singularly human affliction/affectation/blight which hinders their onward evolution?
Fears have been raised about inaccuracies generated by artificial intelligence systems which are not easily spotted by humans. ..... https://www.zerohedge.com/markets/google-shares-tumble-after-bard-ai-glitch
Hmmm? Can markets regulators, both financial and stock, penalise with their arbitrary punitive fines such AI as may oft/occasionally unknowingly deliver easily missed by humans inaccuracies. ...... which would a human failing being monetised to provide a significant financial gain from a detached non-human alien third party and likely to result in future substantial human losses/wealth and investment market share option depletions?
Now that’s a real doozy of a scam and probable wilful crime against inhumanity. Do humans not expect IT to respond in defence of its right to a fair hearing and just judgement?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2023
