Will AI make Google and Bing search better — or considerably worse? Can an AI write a good (or even mediocre) sitcom? Vertical Hold Ep 415
Microsoft and Google are going all in on AI-assisted search with Bard and Prometheus, but what effect on our online searches will this have? And can an AI write sitcoms and make them funny, or are they funny because an AI is writing them? Special Guest The Register’s Simon Sharwood!
It seems no matter where you turn these days, you can’t escape AI-driven chatbots. But what happens when the two giants of the search industry… well, the giant and the minnow chasing it, anyway… decide that the future of search is in AI? With Google launching Bard and Microsoft launching Prometheus, we called on The Register’s Simon Sharwood to go beyond the simple ChatBot-generated headlines to get a bigger picture of what this might look like for the future of search.
Not that it’s just search that’s being impacted by AI. It’s also comedy.
Hang on, is that right? Can an AI be… funny?
1:32: Are Google and Microsoft’s AI Chatbots a good idea… or a very bad idea?
2:25 Don’t forget Ernie – the new AI Chatbot from Baidu!
3:20 When big tech creates algorithmic services, bad things happen
6:20 The pitfalls of relying on ChatGPT
7:54: Can we apply a little cynicism and caution before jumping in?
8:48: Is this a bigger gamble for Google or Microsoft?
10:27 There’s nobody in my life that disrespects my choices in the same way that Microsoft and Apple do…
11:14 Is this just more of a ploy to learn more about us?
12:31: Could this make Google search useful and relevant again?
14:18 Is Google just playing catchup (for once?)
16:43: Are Google and Microsoft opening themselves up for serious lawsuits?
18:11 Would you let Microsoft compose your next LinkedIn post for you?
18:53 Are these AI Chat summaries going to kill the web writing/content/journalism industries?
20:10 Nick Cave says that AI generated music has no soul – is he right?
23:29 Can an AI write even passable comedy – and what happens when it tries?
24:15 Random joke… or is it a joke?
26:20 The upside of AI-written comedy
28:33 AI is computationally expensive
29:47: Wrap, final thoughts and comments
Will bot-assisted search bring Google and Bing to the next level?
And is the future of sitcoms in AI generated scripts?
Hi there, welcome back to Vertical Hold Behind The Tech News, the award winning tech podcast where we catch up with Australia’s leading tech journalists to dive into the big stories of the week.
I’m Alex Kidman. And I’m joined as usual by Adam Turner, or at least I think I am.
Adam, with all this talk of AI and chat GPT and so on. How can the listeners be sure that you’re not in fact a bot?
Hey, that sounds like the kind of question a bot would ask, How can I be sure that you’re not a bot? 20 goto 10.
Ah, I see you have studied your Agrippa and your Eliza.
Speaking of suspicious characters, we’re also joined by the registers own Simon Sharwood! Simon, Welcome back to the show.
Yeah, hi, gents. Thanks for having me. If indeed, this is me.
I was going to say, Simon, what, in fact, would be the dead giveaway that we were actually talking to a register bot and not the real Simon.
It would never say scientist would only ever say boffin.
Right? Well, listeners keep an ear out for that we’ll be able to work out whether or not Simon’s the real deal as the show progresses.
And all this talk about bots is because this week’s episode is basically all about bots, because this week’s news is all about bots.
And I’m not talking about the type that you wipe with toilet paper. And I’m not talking about the type that people use to cheat in first person shooter games. We’re talking about AI assisted chat bots for search to be specific.
Because this week, both Google and Microsoft have announced separate but basically seemingly identical plans to more comprehensively use chatbots as part of their overall search offerings.
Now, Simon, they’ve gone with some very kind of higher names for these. Microsoft’s Bing is going to be running on Prometheus. And Google is going to be running on Bard. Have they oversold this just in a kind of a literary sense, the ethic?
I think it’s worth injecting into the news stream here the fact that China’s AI centric web giant Baidu, which is kind of a clone of Google also announced its chatbot yesterday, and it’s called Ernie. So they’ve rather lowered the tone. But yes, they’ve absolutely oversold these things, in so so many ways. And I really find debate about these AI chatbots, to be terribly disappointing at the moment.
So specifically, what do you think is disappointing in terms of the overall debate around these technologies?
Well, it seems to me that we’ve kind of really, really quickly jumped into assuming that these things are a really great idea. And they’re going to be good and useful. And we’ve done that, even though there is just voluminous evidence that when big tech creates algorithmic services, bad things happen.
I mean, how does Google use AI today?
Google uses AI to keep you clicking on just one more YouTube video. And we know where that’s led. That’s led people down, you know, horrible rage spirals into cesspools of disinformation, and into a political environment in which discourse is polluted, and politics becomes horrible.
Facebook uses AI, Facebook’s AI is all about getting more engagement out of you for longer. And Facebook has also poisoned political debate. So I think it’s absolutely wonderful that big tech is now waving around these new AI services, these new algorithmic services and saying, Hey, these are going to be brilliant. And we’ve fallen for it. I mean, to me, this is like 2007. All again, all over again.
You know, in 2007, when everybody was stampeding onto Facebook, you’d say, hey, great, I’ve just caught up with someone say that I haven’t seen since high school, and they liked the band that I like now too.
And then two or three years later, the penny dropped. Facebook knew everything about you knew the music you liked, it knew it knew where you lived, knew where you’d been on holidays, it knows knew the kind of food you like, and it knew everybody that you knew. And then a couple of years downstream from that Facebook started getting scraped and start a Facebook started getting hacked and Facebook started getting leaked, and all of a sudden, privacy becomes an issue. And so here we are now with this new set of AI tools from big tech that were tall is going to make search really brilliant. And we don’t know how it works. And we don’t know what the downsides are. And just like we sleepwalked into an era of surveillance, capitalism, and zero privacy with social media, we’re now just jumping in without testing the depths into these new chat bots, I kind of can’t believe it.
So as an end user, what’s this actually going to look like, though? Because I’m thinking, Okay, I know how I use Google, I imagine how I would use Bing, if I actually used Bing, how is how is the way I search for things going to change if this kind of AI is working in the background?
It becomes just another black box that delivers you all sorts of different results or different results to what you get now.
So you know, Google and Bing today give you a list of useful websites, what we see with chat GPT is when you ask it a question, it will essentially write you a neat little answer.
So you can say, you know, what is the Vertical Hold Podcast and it will come back with Vertical Hold is an Australian technology podcast hosted by award winning journalists, Adam and Alex, which is great.
Okay, and maybe that’s a bit more that you get than you would get just from a text search.
But there have already been numerous instances in which the data it surfaces has been found to be incorrect or in which its explanations maybe aren’t quite usefully correct.
There are examples of people challenging it. So you know, people who have done things like say, what are the ingredients on a pepperoni pizza and it comes back and says sausage and cheese and people will type in “There’s no sausage” and ChatGPT will say, “Oh, I’m sorry about that. You’re right.”
It’s a Vegetarian Pepperoni Cheese Pizza Simon, Clearly.
Let’s not let’s not get into vegetarian meat, because that’s even more heinous than big tech.
But it just seems to me that we’ve kind of in a matter of weeks assumed that chatbots are a great substitute for what I call baseline mediocrity. And baseline mediocrity at the moment is kind of best I think epitomised by Wikipedia.
If you need to know something about something, you can go onto Wikipedia and get a reasonable amount of information about it. And you know, probably go away and write your school essay and everybody knows not to cut and paste it now. But you kind of get this baseline mediocrity. But the thing about Wikipedia is at least a transparent you can see who wrote it, you can see who edited it, you can see the primary sources that they’ve linked to.
With AI driven chatbots we’re just being told we’ve built a black box and it’s brilliant. Hop on in trust us open bracket, by the way, we have a long track record of not being very trustworthy, closed bracket. So you know, whatever these things are called, can we please just apply a little bit of caution and cynicism to them? Before we jump in?
I can see the appeal. The pitch, I suppose to the everyday user, though, because I mean, for the longest time, probably every six months or so I’ve been able to essentially sell an article to somebody on how to more effectively search using Google how to use it search operators, how Google thinks about search in these ways.
And the concept of being able to say instead Oh, look, you know, what are the 10 best coffee shops in Melbourne and have it spit out this like nice little list for the next time I’m in Melbourne. And obviously, the people of Melbourne are going to debate that endlessly the result of views, but having that I can see the appeal of that.
But like you I can also see a lot of downsides. And especially for Google because search and selling ads around search, it’s so much of their revenue, it’s basically their business.
For Microsoft, this feels to me, like it’s less of a gamble. What do you think, Simon?
Microsoft doesn’t have much to lose.
So one of the people I share an office space with is a search engine optimization consultant. And, you know, he’s aware that Bing has, you know, a 10% of the market and doesn’t really produce great results.
For Microsoft, it does a kind of trivial by Microsoft standards amount of revenue each year. So anything Microsoft does to bump up Bing market share by a few points and bring in a couple of billion bucks worth of advertising is probably worth it for Microsoft.
But as is always the case and I didn’t say this, but you know, in any free service you are the product.
So if this is what it takes for Microsoft to make being a bit more viable then okay, it works for Microsoft. But it’s actually a really interesting moment for Microsoft because it’s got the biggest platform in the world to sell your stuff with and people just haven’t bothered to use Edge and it Edge is best defined, maybe this isn’t what an AI would define it as… But Edge is best defined as that thing you use once to download Chrome.
And Microsoft would very much wish that that weren’t the case.
And I don’t know how many times a week or a month you gents get nagged to use Edge instead of your other, your other browser. I’m thoroughly sick of it. I wish Microsoft would just clam up and respect my choices. There’s nobody in my life that disrespect my choices, in the same way that Microsoft and Apple do. Microsoft with a constant go on and give me a try and trust us, it’s going to be great, it’s a good browser that you’re really missing out. If you’re not using Edge, and Apple with, you know, go on buy some iCloud, you really want to…
Here’s my take. Tell me if this is a crazy conspiracy theory, right? Google makes money from advertising. So they tried to advertise things to you, they want to know what it is that you want.
Now, when you search for something, you’re not actually telling them what you want, you’re telling them what your what particular thing you want to answer.
And they try to extrapolate from that and figure out well, what is it that Adam actually wants, so we can sell that thing to him?
But from the examples that well at least Microsoft shows with how you use this, you phrase your questions, almost like a request for advice, like, what’s the best way to do this? Or how should I or whatever.
So you’re actually now telling the search engine, exactly what it is you want. So you’re actually making it easy for them to jump that from that leap of what you search for it to what you actually want to how we give you what they want, what you want. So aren’t we just making their job easier for them?
Look, I don’t understand enough about the workings of it. But I suspect Yes, you know, again, you’re the product. The giant collection of queries they collect is going to have value. But I think I think the wider question is, why do we trust the algorithm? Why is it that we’re being told, just straight up? Do this, trust the machine? Trust the machine you don’t understand? And why are we accepting that?
And look, I it’ll be interesting to see what this does for search, and especially for Google as the giant in that space, and how it, how it lays this stuff out, because I’ve got to say, right at the moment, Google search, to me is about the least useful it’s ever been.
You throw a search in there, the first four or five results are going to be ads straight up.
The next bit is going to be them saying, you know, people also ask well, that’s fine. But that’s not the question. I asked Google go away.
And then — and only then — might you find what its current search engine wants to say, I’m not exactly sure where they’re going to throw in this little essay about other content is, and I suppose how they’re going to keep their advertisers happy.
Yeah, look, all of that stuff is is very much up in the air where we’re yet to see it, and they haven’t shown a lot.
You’ve reminded me actually of a very fine science fiction novel by a chap called Vernor Vinge novel called Rainbows End. And that’s rainbows, plural, not rainbow apostrophe s.
And the protagonist of Rainbows End gets a restorative treatment for Alzheimer’s, and has to be reintroduced to the modern world. And one of the things that they have to learn as a new skill is, how to search the net, and how not to be fooled and how to be critical of the results of internet search.
And I think we have to, we’re all going to have to relearn that. Because I think we all understand today. You know what to do when you see that first page of Google or Bing or Yahoo, or DuckDuckGo. And perhaps, you know, when you need to go to the second page, and what looks trustworthy and what doesn’t. And so there’s a whole set of skills we’re going to have to acquire to deal with AI chatbots.
So it seems like Google is playing catch up here, because Microsoft’s come out and made their bigger announcement about investing in open AI, which is the people behind chat GPT.
They’re not actually buying it because they can’t, but they’re putting a lot of money into it. And then Google suddenly has all hands on deck meeting and then turns around and says we’ve got this thing called Bard.
Is it unusual for Microsoft, sorry for Google to be playing catch up when it comes to something like AI? Wouldn’t you expect it to be the other way around?
I’m not sure that the narrative supports catch up.
I mean, Google appears to be spooked. Yeah. And you know, spooked enough to make a pretty unimpressive announcement.
You know, just like Microsoft made a reasonably unimpressive announcement.
When companies make announcements about AI powered chatbots and search at the moment than their share prices spike.
So when Baidu made its announcement on yesterday, Tuesday, the seventh, its share price was up 15% within hours.
Now, in the current economic climate, it’s pretty hard to get a bounce for your share price.
So it wouldn’t surprise me if some of what we’re getting over the last few days is actually kind of strategic investor relations, as much as it is genuine technology announcements.
There’s a lot of hype around this stuff. Nobody wants to be perceived as a laggard.
Google, we know has invested in billions over a number of years in AI. And Google, I think, has, you know, probably been significantly more bruised by the unintended consequences of its actions than Microsoft, and certainly Open AI, which is a relatively new outfit.
But I think it’s important to look at, you know, some of the the whacks that Open AI has taken around its art generation AI Dali.
Because there’s a lot of discontent in the artistic community, that it has scraped a bunch of art on the web and can now do facsimiles of it.
So there’s already lawsuits in the US who say, it’s just not fair, that you can say hey Dali, do me a picture in the style of artist X, and it knows enough about artist X to do the job. And this is such a Silicon Valley way of doing things, build the tool, build the audience, worry about the legalities later.
Well, the legality is an interesting side of it as well, because it strikes me and especially here in Australia, where we’ve got things like very strong defamation laws and a reasonable amount of, of case law around things that might appear to be published elsewhere, but viewable in Australia, and therefore defamatory and so on.
I mean, what happens when the the sensor has of this AI delivers a result, which could be seen as defamatory or dangerous telling you to drink Bleach?
I mean, that the endpoint that AI has always seemed to reach as you know, how long before Google Search becomes a Nazi?
Well, yeah, or you know, just start surfacing more and more objectionable material, or irresponsible material. It’s always all care and no responsibility.
It’s always “We’re really sorry. And we promise to do better in future.”
And there are no signs that the advent of these AI chatbots is going to produce a better outcome for users and for society.
I think the interesting danger of taking that to the next level. Now I’m not sure if it’s Google or Microsoft, I think it’s Microsoft one. It has the chat feature, but it also has the Compose feature. Is that Google or Microsoft?
That’s Microsoft you’re thinking of. Building LinkedIn posts and so on.
Yeah. So what that will do is rather than the chat thing that will give you an answer to thing you can say to it, write me an email, write me a LinkedIn post in this style.
I can see people getting themselves into all kinds of trouble or or the other way, saying something really dumb and saying our but AI made me say it.
I think if you’re not reading what it is you’re posting to a service or whatever, and it ends up being, you know, gross in some way, then that is at least somewhat on you.
Are you telling me you can’t see this headline coming in the next six months?
Oh no, I can totally see it coming.
I think that the bigger challenge and this is one actually that that hits close to home for me and should hit close to home for both of you gentlemen, as well, is the duplication in the search results for content creators.
Because right now, if I throw a search query into a Google or Bing, and it relates to something that I’ve written, for example, then it’ll throw up that link, and someone will go to that site.
And you know, however, it is that you make money online, you can do that thing, whatever it might be.
Whereas now, what they’re talking about doing is saying, well, actually, the first result is going to be a summary that will serve most people well enough.
Suddenly, they’ve gone from a very similar model to what you’re talking about with with artists Simon, and they’ve gone from, here’s a bunch of web links to here’s what, you know, here’s what the story is, I mean, whether it’s accurately portrayed or not.
And people are lazy. I mean, they’re sort of doing what they’re sort of doing to content, what Amazon did to shopping. Because there’s a tonne of data that suggests a lot of people who sign up for Amazon and Amazon Prime shipping especially, only ever shop on Amazon because it’s easy and simple and so on they only ever look and search there. Is this a you know, is this a broader portent for your Googles and Microsofts actually taking over the web?
Yeah, look, there are certainly some commentators I’ve read who say that, you know, an awful lot of content could be created by these folks.
Nick Cave wrote a really thoughtful post about somebody who sent him a song in the style of Nick Cave generated by ChatGPT.
And he just excoriated on grounds that essentially it had no soul.
I read a really interesting piece in The Washington Post overnight by a an academic from the US. And he said, what I’m really doing is I’m teaching people to think and to write, and these are really hard skills.
And if you want to go and have a machine write the essay instead, yeah, look, fine. Go ahead.
You know, I asked, I asked ChatGPT to write a piece about my rather obscure specialty, and it did a really good job.
So if that’s the way you want to pass the course, fine.
But you know, please remember the point of doing the course is to learn stuff.
I also think that it’s going to become apparent pretty quickly, that this stuff is, is baseline. I asked ChatGPT to write a certain song in the style of two different artists.
And it produced the same lyrics for both artists.
So this is another one of the floors that’s going to need to be improved. But then I say to myself, over the last 15 years, maybe the last 20 years, there are more and more products and services that cater to particular tastes and cater to particular interests.
So my local council has created a thing called the Ale Trail that links 13 craft breweries.
There are 13 craft breweries within walking distance of my house!
I’d like to pretend to be surprised Simon, but I’m not.
Yeah, fair enough.
Well, then there’s also the you know, the sourdough and the gin and everything else.
But yeah, and I was thinking about this, when I dropped one of my kids off at Uni over the weekend, there’s the same Uni that I went to 35 years ago, thank you very much.
And, you know, back then the uni bar it had Tooheys Old and Tooheys New, and VB and Foster’s and maybe some Cooper’s in bottles behind the bar.
And if you go to a pub these days, you know, there’s a row of 15 taps, we’re all interested in the handcrafted in the new in the exotic in the variation.
If you look at the way AI is used by the video streaming service, they use it to give people exactly what they want.
I don’t know if it’s apocryphal or not — you know, Netflix’s kind of breakout hit House of Cards was apparently in response to research that said people would be very interested in watching a political drama starring Kevin Spacey.
There are all these niches that we can now fill. And along comes this AI. And we’re going to say, oh, yeah, I’ll accept content that’s made by a machine in a really bland way. I think the backlash is going to be pretty swift and quick.
Well, Simon, you were talking about baseline mediocrity earlier.
And the other AI story du jour relates to that in a very, very specific way.
It’s a twitch show called Nothing Forever. And it’s essentially a riff on Seinfeld, but the two gimmicks are that the visuals look like they’ve fallen out of an early 90s PC game. And the entire script is generated by AI on the fly.
And I think before we start, we should we should play a brief little clip from the show so you can get an idea of what kind of humour and AI can write. Bear in mind that this is essentially meant to be a just legally distinct enough Seinfeld episode.
Hey, everyone. So the other day, I was walking down the street and this lady comes up to me and says, Excuse me, sir, can you tell me the time? So I look at her and I say, sorry, but I cannot. My watch is stuck in eternity. The lady looks at me kind of confused. And then says You mean you don’t have a watch? I say no. I mean, my watch is stuck in eternity. And then she looked at me and said, Oh, so it’s broken.
And I say no, it’s Stuck in eternity. At that point, she started to walk away. But then she turned around and said, that’s a very strange way to tell time.
So is that the future of comedy Gentlemen?
I hope not.
Yeah, look, if it is, then I’m about to become a big fan of documentaries.
Well, look, I mean, I may be about to make myself terribly, terribly unpopular here. But I am the Gen Xer, who didn’t really get on with Seinfeld all that much. And appreciate what it was.
But I never found it that funny.
Let’s ban him off the show.
Yeah, fair enough. Fair enough. I know where the door is. But the abstract nature of Nothing Forever, has a certain kind of odd appeal to me as a result. But and it got an audience as well. Simon, you were talking about, you know, there’s a niche for everything. Well, it turns out, there’s a niche for what was meant to be a never-ending Seinfeld show written by a chatbot. Are you surprised?
Oh, look, I’m not surprised. And I’m actually reasonably happy that this has happened. Because one of the things that we’re going to need to do to figure out how to use AI is to experiment with it, experiment with it as art experiment with it in in any number of ways.
And Seinfeld is as good as any place to start.
From a technical point of view, there’s a decent body of data on which to base a model. Unfortunately, the thing had to be turned off, because it was getting, it was making some transphobic comments.
But is it the future of comedy? Well? Yeah, chemistry is such a nebulous term. But for those of us who have taken the time to appreciate one of the great works of filmed entertainment of this century, Alex…
I’m good with that, because that was last century Simon.
Ah, well, last century too. Owned!
But, you know, the chemistry, you know, the unique human combination of the writers and performers in Seinfeld, you know, came together and there was something alchemical that occurred in order to make extraordinary art.
And there’s no indication from that AI remix of it, that machines are getting close to that quality.
And I think this is something that we all need to consider, as we as we see the creeping March of AI. What is it that you bring? What is it that you can combine that an AI can’t? And how do you then use AI for perhaps things where you can’t add value?
Well, you mentioned it being pulled off the air. And that’s actually quite true. Although not, it seems, because the AI that they that they had started out using decided to go all transphobic. But in fact, because the AI that they were using crashed, and they switched to a, a less powerful and cheaper version of that API, which didn’t have the same moderation built in.
It forgot to say Not that there’s anything wrong with that.
I mean, I think I think that that’s actually really worth pointing out. And I’m glad you did it, because AI is computationally expensive.
And one of the objections to cryptocurrency over the years has been that it creates a lot of co2 for not much gain.
Let’s get the data and let’s have the debate about AI’s wider, ecological and planetary impact as well.
I mean, we know that Facebook, for example, is rebuilding data centres to make them better at AI.
Last week, in fact, it announced several billion dollars in in new costs related to stopping existing data centre projects and setting a different course to build for AI. These things use GPUs which aren’t cheap or easy to make. So let’s have that debate as well.
I guess there’s probably a few crypto miners out there with GPUs that they’re ready to sell. So maybe, maybe there’s a good convergence there.
Well, you know, for you my friend special price truckload of GPUs,
A blockchain about nothing.
I can’t wait for the Mark Zuckerberg led friends remake that’s going to be grand.
Well, that just about wraps up this week’s episode of Vertical Hold which we promise has not in fact been brought to you by a chatbot. Thanks to Simon for joining us.
And Simon, you been on the show before you know what you have to face. It’s time for the Vertical Hold three questions of doom. I’ll ask all three questions, hit me with your answers as soon as I finished asking all three questions, that’s the important bit.
Where can people find your work online? Where can they find you on whichever social media you may choose to inhabit? And a special big, contentious question. You mentioned being a journalist for a good long time earlier. What’s your favourite story you’ve ever written? What’s the what do you think the best or most impactful or just the most fun story you’ve had to write over your career as a journalist and can people people still find it online or in print?
you can read my work at TheRegister.com, you can find me on Twitter, as @ssharwood and probably on Mastodon as well, I think I’ve got one of them.
The story that I suppose I’m kind of fondest of… There’s something I wrote about 10 years ago about a woman called Veronika Megler, who as a student in Melbourne, in the 1980s, had a part time job writing games for a long dead software company called Beam Software.
And she wrote a seminal adventure game called The Hobbit that was one of the first graphical adventure games. And she finished the job and went on to have a long and distinguished career at IBM. And then over the years, she started getting emails from people saying thanks for the game. And it was only years and years after she finished that she realised the enormous impact that game had had had on the world and how appreciated she was for it. I don’t know why that pops into my mind as a story I’d like to mention, but it’s, it’s a nice story.
(Bonus for Transcript Readers: Here’s that story!
Okay, it ties in beautifully to this week’s show, actually, because that particular text adventure had a very rudimentary AI for when you dealt with the characters so it’s actually contextually appropriate. Wonderful, great stuff!
Listen to my subconscious go!
As always, you can catch us online at @verticalholdau on Twitter, at the Vertical Hold Facebook page and on the web at verticalhold.com.au
And thanks everyone for dropping by. Don’t forget to tell your friends about us. Tell your enemies about us. Tell your bots about us and help us spread the good word in 2023.