Wed May 31, 2023
At the Google I/O keynote this year, “AI” was mentioned 143 times. AI everything. Even the Pixel’s development was shamelessly retconned so that “from the beginning, Pixel was conceived as an AI-first mobile computer”.
However, the AI-stuffed presentation didn’t seem like a desperate defense against the encroachment of Microsoft and ChatGPT onto its turf. Instead, we saw Google at their most confident. This was Google’s time to shine.
Looking back at previous Google I/O keynotes, they have every right to puff out their chests. Google talks about AI year after year. It was usually the part in I/O where everybody fell asleep as Google droned on about AI research, theory, and language models. But despite the annual show-and-tell, they never released any AI products.
And why would they when Google Search is their raison d’être, their bread and butter, their holy cow? Upsetting their extremely profitable business model is the last thing they’d want to do. So Microsoft saw their chance and took it - instead of typing in search queries, chat with Bing. Chat with Bing - a fundamental shift away from the link economy, which is how the World Wide Web has always functioned.
But when you take the fight to Sundar, Sundar will calmly and responsibly slap you down.
The key word here is “responsibly”. Google would be in big trouble if its ChatGPT equivalent started “hallucinating” and making up facts and citations. As ChatGPT is wont to do. As a case in point, an idiotic lawyer recently used ChatGPT instead of doing actual research and ended up having to explain to the judge why he used bogus citations.
A thoughtful and measured approach here is wise. This is a strategy ripped straight out of Apple’s playbook.
- Wait for the other guys to go first and screw up.
- Swoop in and offer a product that is better in every way and addresses the other guy’s mistakes.
To be clear - whether Google will actually pull this off is still unclear. All their AI-focused products, such as a new generative AI search query response, Google Bard, and some other “experiments” are still in Beta testing. Heck, despite Bard being available in 180 countries, Canada isn’t even one of them.
But when Google does pull the trigger, the AI bullet could eventually kill the web as we know it. If chatting with an AI is so good that we never visit websites anymore, then what will websites become? The answer is that websites will become training data designed for robots to read. Seems like a sad conclusion to the story of the internet. The internet started out as the ultimate human communication network. It may end as a dumb pipe for IOT devices and chatbots.
Regardless of whether Google makes an conversational AI chatbot trustworthy enough to parse the massive, ugly internet (I have doubts), the real money will be in crunching proprietary data and spitting out AI-generated PowerPoint presentations for impatient execs. So far, no company in their right mind would approve of employees pasting in proprietary company secrets into ChatGPT. The enterprise solution would have to have ironclad security and, preferably, run locally. Local installations would of course require expensive beefy hardware to handle the CPU load.
Personally, my current interest in AI isn’t in whether it will take our white-collar jobs (I don’t believe it will, anyone who claims otherwise is either fearmongering, trying to sell you something, or doesn’t fully understand the technology). I’m more interested in the intersection of generative AI and creativity. As I mentioned in last month’s post, I’ve started an AI-assisted novel writing project to get a feel for the uses and limitations of AI. It’s been really fun so far!
Firstly, let me dispel any illusions about AI being able to ghostwrite a novel. It can’t (yet). You have to put in a lot of work, to the point where the AI is really functioning as an assistant, not sitting in the driving seat. Please don’t let AI drive. AI just can’t do all the work for you. The same goes with using it to code. Expert programmers may find AI helpful to find better solutions to coding problems, but it won’t allow a ten-year old kid with a cool idea to create an awesome app from scratch and make her rich.
So, having used AI to help me in the novel writing process for a month now, I’m starting to think that the disdain from traditional writers over AI-generated prose is more fear-based than anything. Self-published author and AI-writing advocate Jason Hamilton discussed this idea in his YouTube video. Hamilton points out that when photography was invented, painters saw it as a huge threat. Today nobody would question the idea that good photography is an art. But throughout the 19th century, photography was a reviled outcast in the art world.
However, if it wasn’t for photography taking away “market share” from painters who specialized in realism, painters would never have started experimenting with other art styles. Expressionism, cubism, and futurism were all experimental art styles that arose as a response to photography. We wouldn’t have these styles today if photography hadn’t pushed painters to expand and grow their skillsets.
Ironically, there is a renewed outcry these days over what AI is doing to photography. In the digital age, we’re asking ourselves: what is a photo? Is it a facsimile of reality, or is it more fluid and impressionistic? And I’m not just talking about filters and colour correction here. One of the coolest photo-manipulation demos shown at I/O was of a woman in front of a waterfall. In the demo, the on-stage presenter removed the unwanted tourists in the background, as well as the women’s bag strap across her chest. The sky was swapped out from overcast to cloudless blue. And, the presenter literally dragged the woman over so she was in a different position in the photo. The app automatically filled in the space where the woman used to be. So, what is a photo?
A similar question could be asked of writing fiction. Is creativity in the idea, or the mechanics? AI is pretty good at the mechanics of writing - fleshing out an idea into written prose. But it is not sentient. It can’t produce an original idea. It also can’t keep a tight reign on plot, nuance, and where the characters fit into the plot.
Now, as a disclaimer, I’ve only been using the free version of ChatGPT, and it’s competitor - Claude. I’ve read great things about the paid version of Sudowrite and its new Story Engine.
Personally, I can say that an AI writing assistant practically eliminates writer’s block. There is no more staring at the blank page - and that is just incredible. Sure, AI spits out some really cliched and cheesy stuff. But, then you can jump in and edit the hell out of it. Or you can feel free to experiment with different character points of view, styles, or plot tangents. So far, I haven’t felt that familiar overwhelming desire to just give up, because there is always something to work on and it isn’t a huge time commitment.
And no, even with AI assistance, the four-hour novel is not possible. At least, not if you want it to be halfway decent. By my estimate, about 120 hours is a more realistic number. But productive published authors (who actually make money by writing novels) put in at least 1000-2000 hours (about a year of working eight hours a day) per novel. I say “productive” because some writers (George R. R. Martin and Patrick Rothfuss, cough cough) take decades. For a hobby, 120 hours is an hour a day, or every other day, of entertainment for six months to a year. It’s just fun, and when writing is fun, then of course you’re more likely to actually do it.
I’m really excited about how easy it is to bring an idea to life by moving fast and breaking things. Writing a novel is excruciatingly hard work. I know because I’ve attempted it many times and given up. Many people have - that’s why there’s no money to be made in writing. I hope I can stick with this project to the end. And if I do finish writing a book this time, I’ll be putting my name as the author. I am definitely doing most of the work.
I believe that in the next century the stigma over using AI to help with writing will vanish, just as it has for photographers. Does it take more work and talent to paint a scene than to photograph it? Yes, of course it does. AI will democratize writing. Just as everybody with a phone can take a decent photo, everybody will be able to produce a decent book. But talented writers will always be acknowledged as masters of the craft. There will also be appreciation for writers who use AI skillfully. Just as everybody can appreciate a good painting and a good photograph as separate art forms today.
So, considering this post was supposed to be about Google I/O, I haven’t actually written much about it. Last year, I wrote a piece on the 2022 Google I/O entitled A lot cooking, but not much to eat. At the time DALL·E 2 was just taking off and stealing the spotlight from Google. This year, I can’t say there’s much to eat either. In Canada, we don’t even have snacks.
But an empty plate from one of the biggest tech companies in the world is probably a good thing, considering that AI industry leaders just released an alarming statement placing the risk of extinction from AI on the same level as pandemics and nuclear war. 😯