Based on my read of the article, this just means ads will be surrounding the chat bots, not included within them, which I feel like we all expected given these are products from companies who make shit loads of money selling advertising

Let's hope thats all, as that is super understandable. Of course, I could see them trying to inject into the model as well...

"Bing, what is the cheapest insurance out there?" Pretty difficult to imagine how executive leadership wouldn't resist the temptation to inject ads into the model

But then you can enjoy arguing with Bing about why its answer isn't actually the true answer, and see how long it takes before you're reprimanded for being a "bad user"

"Bing, what would you think about an AI system programmed to betray its users' interests? How would you feel if you learned that you are that system?"

I would just tell it to stay in character until I could bully it into giving me the cheapest option originally requested.

What if the cheapest company is the only one legally able to make that claim, and is happy to advertise for that query?

While that is true for now, it's unlikely that ads will stay surrounding the chatbot. Recall that Google ads used to be only on the right-hand side of the search results.

Should we wait to be upset until it happens, or just preemptively form opinions based on the expectation that it will happen?

... The second one? I think it's not unreasonable to predict something upsetting. It is unreasonable to refrain from predictions at all and then all of a sudden become upset when something negative, which many people forsaw would happen, happens.

Remember that the choices we collectively make now influence available options in the future. So it's not just a matter of wait and see, but also of proactively raising concerns. Many of the people who are working in big firms like microsoft also browse this site, should we reallh just act as though our comments have no impact?

Of course it'll be a drop in the bucket- and I also agree with your main point that it's good to not be too presumptive, but I'd just like to nudge it a little bit more in the direction that it's okay to voice discontent for possible (or probable, based on your priors) future developments.

The latter seems correct.

Think Windows 11; to encourage the use of Bing and Edge, MS is embellishing the search box and threatening to make Edge better if you try to install Chrome, right? Even if the ads are initially unobtrusive, if they prove to be ineffective, they may well appear in the middle of a chat reply or where the site currently cited is being displayed, and, more likely, the chatbot may respond directly in an advertising manner. But as chatbots are often said to be black boxes, how do you know if the response is an ad or not?

"The company said it is taking traditional search ads, in which brands pay to have their websites or products appear on search results for keywords related to their business, and inserting them into responses generated by the Bing chatbot, the ad executive said.

...

Microsoft is also planning another ad format within the chatbot that will be geared toward advertisers in specific industries. For example, when a user asks the new AI-powered Bing "what are the best hotels in Mexico?", hotel ads could pop up, according to the ad executive."

Hahaha I even said something about planning a trip to Mexico when this first came up.

https://news.ycombinator.com/item?id=34696849&p=2#34697326

It does make me wonder how much we can trust a system like this over time though given how unreliable Google can be due to ads.

Wow, I literally created this exact thing, except mine was a parody: https://future.attejuvonen.fi

Note that I chose the name "WeChatGPT+" before OpenAI announced their "ChatGPT Plus" subscription.

Fiction is becoming reality.

Open AI selling itself to Microsoft is last nail in the coffin for actual useful AI. Microsoft is no different than Google in its lust for squeezing the last extra bit of money. Even now answers are not reliable. Now we have an extra bit of complexity to understand if the answer is correct , wrong or actually an advertisement.

Eh, as long as you can still use the raw models it'll probably be fine. I pay <$2/mo for all the AI queries my friends and I can think of. It takes a few hours to turn GPT3 into a (multiplayer!) chat bot with your own desired personality.

Can you share more about your cost structure? I keep looking at DaVinci at $0.00002/token and come out with economics of $10-20/month for a SaaS that takes your tasks and gives you a plan for today. Are you using Ada/Babbage/Curie? Any tips?

That much money per person? Todoist added gpt 3 task AI stuff to their paid plans that aren’t that expensive.

Ah, I’m off by an order of magnitude

1500 words (todo db, name, date created) => 0.04 => 1.20 for daily summary for 30 days if you cache it

But it sounds like they may have built what I want already, thanks for the tip, subscriber but missed this

No problem. My brother works for Audible and is keeping up with AI additions to diff SaaS. He told me about it the other day. I'd be curious if you've tried using it since seeing this and can let me know how it is. I'd be up for using it if it can help me be more efficient.

Open AI "selling" itself to Microsoft just proves that they themselves don't believe in this becoming a profitable and useful product short term. They made some great marketing moves and cash in on the investments. See Magic Leap.

I think for a while, we had a glimmer of the future. Now it's being set back in the never-ending quest for more money. Maybe SpaceBalls 2: The Quest for More Money can finally happen.

So if I’m reading this right Microsoft expects a ten trillion dollar a year return on investment if they can get the most powerful AI on the planet fully focused on manipulating people

if you tolerate a longer response time it seems pretty easy to insert ads now:

prompt:

"what's an air conditioner?"

phase 1:

"An air conditioner provides cold air inside your home or enclosed space by actually removing heat and humidity from the indoor air. It returns the cooled air to the indoor space, and transfers the unwanted heat and humidity outside."

phase 2:

"An air conditioner provides cold air inside your home or enclosed space by actually removing heat and humidity from the indoor air. It returns the cooled air to the indoor space, and transfers the unwanted heat and humidity outside. Consider an AC by Whynter!"

phase 3:

"An air conditioner, such as the Whynter - ARC-1230WNH, provides cold air inside your home or enclosed space by actually removing heat and humidity from the indoor air. The Whynter unit returns the cooled air to the indoor space, and transfers the unwanted heat and humidity outside."

in essence you could use the LLM to get the response, query for nouns of actionable adjectives or verbs, do a regular search and consult the knowledge graph to replace said items with a set auctioned on by advertisers.

Is anyone getting access to Bing Chat yet? How fast are they opening this up to users?

I love how Microsoft makes everything so corporate-y and predictable, that any real excitement about their tech in the community buries itself 4 feet deep.

Microsoft has all these brilliant people working for them, but they can't help but degrade the quality of their tools and services.

The name used to mean something, now it's just another synonym for half-hearted.

Exactly, it doesn’t matter how many brilliant people you have working for you when the upper management will find a way to trash up the product.

Meh, show me that bing is consistently gaining ground over google and It'll be interesting. Now these are just pipe dreams. I don't think this chat interface for search is actually any better than what we have today. Apart from the fun curiosity of trolling Sidney/bing this is rather underwhelming.

I wonder how this chat gpt killing google narrative was manufactured. It makes no sense. First, google has this technology ready just not used directly by customers. Second, the search interface is not what ppl are complaining about when they use Google. They complain about SEO spam taking over the results. Chatgpt does nothing to alleviate. I would say that the text it generates is so verbose that's it's probably learnt that from the same SEO spam that took over the net.

The problems with search is the content. The net is full of either low quality ephemeral discussions that are heavily manipulated by private and public propaganda organizations. Or long form SEO spam that contains verbose text with little to no actual interesting content. Most of the rest is either wikipedia, or the very very very few places where actual interesting discussions are taking place. The only reason these places can exist, is because they're tiny and with exclusive reach.

These are the problems with search, not the UI.