The future of artificial intelligence is boring
Searching for Answers to Problems of Geoscience and Artificial Intelligence: A Response to Hinton, Sundar Pichai, and Hinton
The capabilities of ChatGPT and AI language algorithms like those powering it are so striking that some experts, including Geoffrey Hinton, a pioneering researcher who recently left Google, have felt compelled to warn that we might be building systems that we will someday struggle to control. Openai’s robot is great at generating text on a given topic, and even answering tricky questions, which requires expert knowledge.
And yet, unfettered AI language models are also silver-tongued agents of chaos. They will gladly fabricate facts, express unpleasant biases, and say unpleasant or disturbing things with the right prompting. Microsoft was forced to limit the capabilities of Bing chat shortly after launch to avoid such embarrassing misbehavior, in part because its bot divulged its secret codename—Sydney—and accused a New York Times columnist of not loving his spouse.
generative chatbots is a wild side of the internet that needs to be reined in. But in my tests, the new search interface felt incredibly tame compared to ChatGPT or Google’s own chatbot Bard.
It is not possible to use the first person or talk about thoughts in the smarter version of search. It avoids topics that could be considered risky and doesn’t offer answers to controversial topics such as US politics.
In March, some big names in AI research signed an open letter calling for a six-month pause on creating machine learning systems more powerful than GPT-4, which powers ChatGPT. Pichai said in his keynote speech yesterday that the company is currently training a new, more powerful language model called ‘Gemeintank’, which was not a signatory.
A source has told me that a new system may eclipse GPT-4 because of recent advances from different large language models. But don’t expect to get to experience the full power or charisma Gemini can offer. If Google applies the same chaos-taming methods seen in its chat-like search experiment, it may just seem like another surprisingly clever autocomplete.
Sundar is performing that now. He just reorganized Google and Alphabet’s AI teams, moving a company called DeepMind inside Google and merging it with the Google Brain AI group to form a new unit called Google DeepMind. Y’all know I can’t resist an org chart question, so we talked about why he made that decision and how he executed it.
The web search is a little bit hit or miss right now. Remaking the web and remaking the search engine, is something that’s been done before, but that’s something that’s going to need to be done again, because there’s a lot of weird content out there, and artificial intelligence may just be able to answer questions
This is a jam-packed episode. Sundar and I talked about a lot, and I didn’t even get to Google’s AI metadata plans or what’s going on with RCS and Android. Maybe next time.
At a few months ago, I was at the launch of Bing. I saw him there. And I’m sure you know this, but
he said
, “I have a lot of respect for Sundar and his team, but I want Google to dance.” And then he said, “I want people to know that Microsoft made them dance.” I want to know how you felt when he said that. Do you think you danced? Are you doing a dance?
For me, maybe I’ll say it this way. We worked on a Search Generative Experience last year. To me, it’s important in these moments to separate the signal from the noise. I think there is a new way to make search better and we need to make our user experience better, but we have to get it right. And to me, that’s the North Star. That’s the signal. The rest is not interesting to me. So to me, it was just important to work and get it right, and that’s what we’ve been focused on.
There is another challenge for Google inside of all this. If you believe it’s a platform shift, this might be the first platform shift that regulators understand because it’s very obvious what kind of labor will be displaced. I gather lawyers, mostly. They can see that a lot of white- collar workers are going to go away, like an email about a transaction. And they seem very focused on that risk. The general Artificial Intelligence risk that we all talk about.
When Google first did Search, it was an underdog, right? It won many court cases along the way to build the internet, including the Google Books case and Perfect 10 case. It was obviously delivering a great deal of value. Now, you’re at the White House having an AI summit. I’m confident you’re going to end up in government capitals around the world talking about AI. Do you think you’re in a different position now than that scrappy underdog inventing the internet? You are the incumbent. Are you in a different role?
Two parts to the question. I think we can on the first part. For the past two decades, people have predicted that jobs would disappear from tech automation. Movie theaters were supposed to end and–
The unemployment rate over the last two decades of tech automation hasn’t fully gone down. Twenty years ago, when people exactly predicted what tech automation would do, there are very specific pronouncements of entire job categories which would go away. That hasn’t fully played out. So I think there’s a chance that AI may actually… Because I think the legal profession is a lot more than… It is possible you know more about being a lawyer. I don’t know a lot about it, so I won’t opine on it. More people may become lawyers because the underlying reasons why legal systems exist are still going to be with us some day. There are some consequences to the use of Artificial Intelligence, and I think it will improve the profession in certain ways, but maybe there will be more lawyers in 10 years.
I think the answers for each of this are not obvious to me. I think it is not a good idea to hold back the machine intelligence in a straightforward way. It has political implications. It is, again, a complex thing that we will have to wrestle with over time. I think, from our standpoint, we are a bigger company, so I do think we will come to it in a more responsible way. We will try to find the right answers in certain places. And so maybe our approach will be different for sure, I think, as we go through it.
Podcasts: Podcasts app for iPad, iPhone, and iPad – listen to podcasts with GadgetLab on Twitter and RSS Feed
If you are on the iPad, there is an app called Podcasts that you can open. You can also find overcast or pocket casts on the internet. You can find us in the Podcasts section of the app, if you are on the phone. We are on the platform as well. If you really need it, here’s the RSS feed.
You can always listen to this week’s podcast through the audio player on this page, but if you want to subscribe for free to get every episode, here’s how:
Julian Chokkattu can be found on Twitter @JulianChokkattu. LaurenGoode is the same person as Lauren Goode. Michael Calore is @snackfight. Bling the main hotline at @GadgetLab. He is the producer of the show. Our theme music is by Solar Keys.
Source: https://www.wired.com/story/gadget-lab-podcast-597/
A Traveller’s Guide to the Top Ten YZ’s and Thermostats (by Julian Malcom and Mark Malcom), Part II: Reflect Aeroearbuds
Julian recommends going on vacation and also the new Legend of Zelda game. The book Still Pictures by Janet Malcom is recommended by Lauren. Mike recommends the JBL Reflect Aero earbuds.
