Google integrates Chat with Gemini option silently into Chrome
It is probably only a matter of time before all Chrome users gain access to Gemini directly in the browser. Microsoft led the way with its integration of Copilot in Microsoft Edge.
While Google's AI Gemini is not yet integrated visibly in Google Chrome, Chrome users may enable the functionality already. The integration that is already available is hidden at the moment, but it may be activated in development versions already. Chrome users may then interact with the AI directly from the Google Chrome address bar.
Whether that is desirable or not is for the individual user to decide. This is an optional feature for now. The company announced Gemini 1.5 today, promising a model with extended capabilities and processing powers.
Note: Chat with Gemini is not yet available in Chrome Stable.
How to enable Chat with Gemini in Google Chrome
It takes just a few clicks to enable the AI integration in the Chrome web browser. I tested this in the latest Chrome Canary version (123.0.6304.0).
- Open Google Chrome Canary and make sure that the browser is up to date. Load chrome://settings/help to make sure of that.
- Load chrome://flags/#omnibox-starter-pack-expansion in the address bar afterwards.
- Set the feature Expansion pack for the Site search starter pack to Enabled.
- Restart Google Chrome.
This is all that it takes to enable Chat with Gemini in Google Chrome.
Using Chat with Gemini in Google Chrome
Now that the feature is enabled, you may start interacting with the AI from Chrome's address bar. To interact with the AI, type @ in the address bar.
You see several options, including three specific search options. The fourth option is Chat with Gemini, which you may select with a click or tap.
When you do that, Chrome displays a Chat with Gemini label in the address bar. Type or paste text into the address field now and press the Enter-key to send the data to the AI.
Google Chrome loads the Gemini website. If you are not signed in, you will see the sign-in option and no response to your request.
If you are signed-in, you may still not get a request, but the Gemini Chat startpage instead. It is possible that Google is still working on the integration.
One interesting aspect of this integration is the path that Google uses to send data to Gemini: https://gemini.google.com/app?q=TEXT
This allows other services, in theory at least, to send input to Gemini right away as well.
Google's Gemini integration in Chrome
Google's integration of Gemini in Chrome has several reasons. The first is that it will give Gemini a lot of exposure. While Google owns popular websites and services, Chrome is ideal to getting the word out about Gemini.
For Chrome users, Gemini's integration may either be useful, if they plan to interact with the AI, or not useful at all. There is a chance that private data gets submitted to Gemini, which is probably less likely if you visit the Gemini website directly.
Companies like Microsoft and Google will continue to push their AI products into their services and applications. Mozilla is also working on integrating trustworthy AI into Firefox, but it will take quite some time before anything of substance comes out of the recently announced pivot in development priorities.
AI will be an integral part of web browsers.. Users who dislike that may get options to turn this off, but there is no guarantee that all browser manufacturers will add these control options to their browsers.
Now You: do you plan to use AI in your favorite browser?
My energy provider uses one of these and it’s totally useless. Regardless of the topic you get a choice out of five options. If none of those even vaguely answers your question the stock response is “Can I help you with anything else?” after which the whole scenario repeats itself.
They have a phone number to call to speak to a real person, but the queue is usually at least an hour long during which time you have to listen to all the wonderful things they have to offer on their site followed by a suggestion to try out their useless chatbot if you can’t find what you’re looking for. I despair…..
Was a spyware, now even bigger spyware
Day #whatever-long-number of still not caring about AI and still having no desire to use it or to have it in my life. People like me are marginalised it seems :/
More of us that have zero interest in AI than you expect. In fact, I have found during recent conversations with peers that the more someone understands how AI functions, the less likely they are to want it anywhere near their systems.
Don’t let the clearly transparent media push for AI discourage you. The only people I have met who are happy about the current hype-cycle are the new-thingy zealots who think it saves time and non-technical managers who think they might be able to save a buck or fear getting left behind.
If you happen to find a zealot who thinks this AI garbage is all fun and games, you can point them to the AI Delimma video on youtube. Might be enough to wake them up.
“Do you plan to use AI in your favorite browser?”
Yes, I use AI for certain searches; it can save time. The results, however, are unpredictable. I may find a treasure of sites I would have missed; or I may be presented with a bunch of garbage.
Note: The use of “already” in the second paragraph. I have “already” lost count of the number of times it appears.
The output is not “unpredictable” but rather non-deterministic. That’s a result of how these neural networks are trained (and initially seeded – with random numbers).
There is a a significant difference between a process being unpredictable and being non-deterministic. Unpredictable processes are random or seemingly random (which is in nature usually caused by non-linearity or a large number of undefined/unknown effectors) and generally can’t be predicted. In contrast: non-deterministic processes are often still statistically predictable (-> this means: ask just often enough (lim n -> very large number) and the output will sooner or later tend to center around the same/or very similar outputs)
Ok given, if you are just a user then for you, there is no difference, but in principle there is one.
Regarding “false” answers:
as for all IT systems: GIGO (garbage in -> garbage out). That’s true for every IT system. So every IT system that is fed with “false” (whatever that might exactly mean) information will also generate false output. Eg the same is true for search engines.
BTW, if you are using a chatbot for information retrieval, instead of a search engine, this just means you are using a chatbot to evaluate internet content search results for you, as you trust the system more, than your own mind (this says a lot about either your opinion of your own mind or of your opinion about the “intelligence” of a Large-Language Model). This use is heavily flawed.
You are right, some of the results of the AI are certainly unpredictable or directly false.
Not just false, dangerously false.
I highly recommend reading up on behavioral economics. Predictably Irrational by Ariely and The Psychology of Thought Control by Meerloo are a reasonable start.
Once you have a foundation in the material, consider the implications Gen AI brings to the table.
It should scare the hell out of you.
IMHO All this AI integration should be voluntary and needs to be toggleble on and off. It is still to early to tell what the security and privacy issues are.
where is the problem ?
“Integration” means: you CAN send data directly to Gemini, nothing more.
You CAN, means: you CAN, but you DON’T HAVE to. You even have to actively prefix the stuff with an @ to do so.
So if you don’t like it -> don’t use it. That’s as voluntary, as voluntary gets.
Further is the AI executed on Google’s servers. You just send data, if you want to and get a response back, just like from any website, so where is the issue ?
Complaining about a company offering a free remotely executed service, that users can use, but don’t have to is as nonsensical and surreal as it gets.
And regarding “risks”, that’s mostly from people with an (at best) limited technical understanding or from avid scifi fans who can’t distinguish a pretty dumb LLM (reality) from an Asimovian positronic brain (sci-fi) or rather Skynet (scifi). What’s called “AI” is not AI in the sense of intelligence, but it’s ML, so a data trained algorithm. That’s all, automated statistics.
Idiot’s fantasizing about AGI simply don’t understand what they are talking about.
If all this LLM stuff is sensible, is a completely different question (especially for those who know Eliza). Imho it’s mostly hype, although it will have some useful applications, eg instantaneous speech translation, text summaries etc.
The less the technological understanding, the greater the fear. Has always been like that. Was already like that in the late 70s, when computers first moved into people’s homes and corporations and people cried and wanted to put all computer use under governmental surveillance, or in the 80s, when “gene technology” made people shit their pants because of all the unmanageable risks, that will undoubtedly doom us.
Security: Fear mongering sells, nearly as good as sex. So there are always tons of people interested in creating a feeling of general threat, in order to profit from it. And those not knowing what these are talking about and not willing to get a deeper understanding about they topic (because learning makes headaches) buy it.
And regarding privacy: AI as a technology itself has no privacy issue, whatsoever. The privacy issue lies with the data being used to train the systems, but that’s not an AI issue, but a problem of the data being used. It’s thus not an AI problem, but a data problem and thus a matter of improper data protection.