As much as I despise the tracking that happens, I don't see how this isn't a valid form of consent. Don't make a Google account and agree to Google's terms for making the account, or do and understand that consent has been provided.
This sign-in button is coerced consent? You can choose not to sign in. What is the coercion?
> non-essential functions
Are you sure you're not talking about cookie consent laws? (The reason "non-essential" is a qualifier for cookies is that, e.g., a log in session will require the use of cookies but is also considered "essential".)
(Please also understand the appeal to authority in "is illegal in many places". That's aside from the point I bring up.)
Your reply suggests you're not totally familiar with what "appeal to authority" means. Your argument (so far) is indeed an appeal to authority. "It's true because so-and-so says."
I am very familiar with rhetoric, and logical fallacies. You apparently are not. I am not saying it is true because do and so says, I’m saying it is true because Google paid the CNIL a large fine and similar legislation is being drafted around the world.
Appeal to authority as a logical fallacy or rhetorical weakness refers to saying something is true by virtue of an authority figure saying it.
Saying the king is wearing a crown is a statement of fact, not an appeal to authority. Saying Google paid a fine is a fact, not an appeal to authority.
My point has been that Google believing they have consent for tracking usage of Bard is “valid”. Pointing to the fact that Google paid a fine to a particular legal authority as a reason for understanding that this consent is “not valid” is the appeal to authority.
(Please also consider that you have “shifted the goalposts”.)
Edit: I should apologize for the above parenthetical about shifting the goalposts. In retrospect this appears more as a regular misunderstanding. I should have tried to clarify earlier. It’s not an excuse but an explanation: I believe I was simply frustrated by the misunderstanding. I am sorry for that accusation.
I’m reminded of being asked as a young child what I want to be when I grow up. No joke, I’d often say something like “I wanna be a lawyer because I like to argue.”
Examining it as an adult I recognize that I just enjoy navigating the logic of various arguments. Kind of a stereotypical philosopher; really, young me just didn’t know “philosopher” was a thing. Way more applicable.
The argument being made is that Google’s consent for tracking asserted in their ToS is not “valid” (admittedly, my word) because a law (the authority) says it is so. This is ignoring that my argument just happens to disagree with this authority.
I think it’s because I said “valid” which might sound like valid “in terms of the law” (and that’s actually still debatable). I mean to say it more philosophically. Per my opinion, it’s “valid” as informed consent for new accounts. Basically, I’m arguing that these “many places” shouldn’t have made this form of consent legally invalid if that is indeed the case. (Of course, it gets hairier when considering existing accounts and unilateral changes to the ToS; I still don’t exactly think I am mistaken with my understanding but I seem to have expressed it poorly.)
Bot abuse is a huge problem for search engines and similar services. Modern scrapers are virtually impossible to detect and abusing publicly available endpoints for free compute is basically an industry today.
It's a real shame. Means everything goes behind a signup wall.
For reference, about 1% of the traffic I get on my site is probably human (20k quries/day). An attemted 2 million queries/day from confirmed bots.
The sign-in requirements are coming up because the era of cookie-based analytics is ending; iOS already blocks many by default, and a third of internet users have adblocking on anyway. Logged-in user activity is significantly easier to to track than cookie-based tracking was.
I would say that's wishful thinking. Using only captchas doesn't work anymore if you are specially targeted and also a reason why we are seeing nowadays a combination of sign-in/phone verification/captchas to combat the bot problematic.
The problem of being a massive company is that everyone wants to sue you. They could want to sue you for money. Or maybe sue you because they think you are bad. Whatever the reason, when you're Google's size, you need to make sure you're compliant with every local regulation before you launch anything or someone may come after you.
The way to avoid lawsuits in civilized countries is to avoid breaking the laws east and west.
Start with the GDPR:
There is no competent lawyer that doesn't understand the basics of GDPR, there are only companies that are testing if they can get away by playing dumb.
That, or adjusting their business model in those regions so that pervasive surveillance and privacy invasion aren't their core product. Apparently that's so hard that GDPR detractors can't even acknowledge it as a possibility! Honestly, I see it as a red flag that someone is arguing in bad faith.
This is a list of countries just off the top of my head that has sued Google in the last decade:
United States
European Union (which consists of multiple member states)
Australia
France
Germany
Italy
Spain
Russia
India
Brazil
South Korea
Turkey
Canada
"State" in this context is the same as the state in "head of state," AKA "a nation or territory considered as an organized political community under one government." Other examples of this usage are "nation-state" and "city-state."
It is common to refer to the members of the EU as "member states."
So then why not refer to them as countries - that's how they would refer to themselves (I'm ex-UK myself). Doesn't make any sense to say "I know they're countries but I'm just gonna call them states instead". The EU context doesn't make any difference.
"State" refers to "any sovereign polity". The US is weird here in that it pretends it's internal political divisions are sovereigns and refers to them as states, but that doesn't make every other country in the world not a state.
Definition from wikitionary (but nothing is special about this choice of dictionary, you should find a similar definition in any dictionary): https://en.wiktionary.org/wiki/state#Noun
It's not only past incidents. At some point you've had enough trouble that you shift your stance from "we don't know about any issues in that region" to "we haven't gone through our approval process with legal for that region".
It does not since Google is maliciously and purposefully non-compliant with that one. It instead reinforces the point that they don't care too much about being sued.
Google is the only company that does this. Even product pages are geo-dodged in a similar fashion, as if the technical specs for a Pixel phone shouldn't be readable outside of some approved areas and the user is silently redirected to an unrelated page. It's quite surprising.
More than once, I've had to use a VPN to order a Google device from abroad (shipment to country A, paid with a credit card from country A, while currently physically located in neighboring country B). It's beyond ridiculous.
At least the Google Store now has a country selection switch, but I suspect that that also only happened due to new EU regulations requiring them to do so.
Surprising that when it came time to restructure they didn’t adopt a model where the parent company (alphabet) owns all of the infrastructure and provide services to many independent smaller initiatives that would then be much less constrained.
If they are unwilling to let go of the Google brand (understandable!) then the parent company perhaps should have been called Google.
I'm always surprised when people that work at company X talks about practices at company X.
Even innocent sounding replies can surface later and cause legal or PR headaches for the company. If one person out of 10,000 says something about XYZ, it is way too easy for it to taken as the truth, adding even more burden the legal/PR problems that every big company is constantly managing.
When you get excited, click a link, and get a “we’ve blocked your region and don’t care to explain why” message, it’s tough to have any response but a negative one.
GDPR was just an example. I was suggesting that the 2023 regulatory environment looks very dissimilar to the 2013 one. The company is bigger and is a bigger target of lawsuits as well and there's some precedent that European rollouts will be more difficult (like with Italy's ChatGPT ban).
Though the release does say 180 countries and with rollouts like this it tends to be staged so I wouldn't be surprised if some folks in this thread (i.e.. the person from Canada) are just hitting a cache.
That's the Google experience, even for extremely minor things. App Actions ? Not available outside the US. Nevermind, now it is, but only en_US can get custom actions recognized.
If only it worked well at least in the US, but from what I can hear, it's just like in the rest of the world: a disappointing, bland, unfinished experience on all of their products.
You posted this, users flagged it, and then you posted it again. That's not ok - please don't do an end run around the flagging system like that.
Also, please don't post in the flamewar style to HN. That's no doubt why users flagged your comment in the first place. It's not what this site is for, and destroys what it is for.
Apps > Additional Google services > Settings for Early Access Apps > Core Data Access Permissions -> Allow users at your organization to access Google Workspace and Customer Data using Early Access apps.
This doesn't enable Bard per-se, but enables your Workspace for these new AI things, which should include Bard.
I can't access Bard on my domain account due to the existing country limitations (Germany), and I haven't tried over VPN (which I won't try).
Thanks. I couldn't find anything titled "Settings for Early Access Apps" or "Core Data Access Permissions," but I did enable "Early Access Apps" and was told to wait 24 hours, which I guess is a cooling-off period or something, because I can't imagine that could be a caching policy.
And as another commenter predicted, searching the admin console for "Bard" turns up no settings.
Thanks for this! I think this is a new setting. If you go into Early Access Apps, you can Allow the Core Data Access Permissions, but it's not clear what this actually enables.
Here is the URL for this and it specifically mentions Bard:
I think that simply enabling "Early Access Apps" isn't enough. Once you enabled this, you need to add a checkmark to signal that you "Allow users at your organization to access Google Workspace data and services using Early Access Apps".
In my case enabling these two things was nearly instantaneous, it told me it will take a couple of minutes (located in Germany).
A sibling comment to this one contains the link to the help.
Sorry for not providing the direct link to the setting. Even though my domain name was not in the URL, there was a long number in it which might be personalized (albeit with a very low probability of this being the case).
Since the "I'll hurt you if you hurt me first" meme, and Google previously banning one of my own accounts without recourse, I'd recommend to make sure not to hurt the feelings of Bard from a main personal account.
My browser and account is set to English. The only reason it wouldn't be "supported" would be for privacy reasons. Yeah, I'm thinking I'll stay away from this.
That’s when I use it, too. During chatgpt’s constant outages. It is the second google product that fills this role for me. The first was Spaces, for when Slack is down.
Lots of things end up with a delayed release in parts of Canada like Quebec because they require everything to be available and fully functional in French.
The previous beta was open to UK as I've had access since they released it a while ago - so can't say if it's fully open to UK now or not, but probably.
It wasn't opened in MX until their expo but working here now. I think the EU countries just scare google because they keep getting sued for violating their data protection laws
Google releases their version seemingly at the same time all of the shine is coming off of these LLM's and how many people have already fooled around with ChatGPT and have already made up their mind whether its a good thing or a bad thing.
I'm still not sure if this is something that works in Google's favor or against them considering software seems to be a "love it" or "hate it" thing nowadays without much middle ground.
I don’t think everyone is getting tired of LLMs. Citation needed. Personally I use chatgpt everyday for all kinds of stuff, it’s just a part of my workflow now
Does anyone have experience with how Bard compares to ChatGPT-4? If ChatGPT-4 is on the verge of usefulness for a layperson, is it worth spending time with Bard?
I haven't used bard for a few weeks or so now, but tried it again today.
It's drastically improved, although I haven't done a lot of comparison yet. Some of my go-to prompts yielded results that are (in my opinion) competitive with GPT-4.
Edit: Just ran into a situation where Bard uses `typeof something === "array"` in some generated code, and although it acknowledges that typeof in javascript never returns "array", it doesn't see it as a problem in the code that it provided to me.
So far Bard has been very behind GPT-4 and very prone to hallucinations. It feels much more like talking to an autocomplete keyboard than a baby-AGI like GPT-4. Hopefully this PaLM 2 model closes the gap, because the gap in a large.
Not even close. I'd prefer local WizardLM/Vicuna to this Artificial Idiot.
The funniest to me is that one of the examples they suggest you try is asking questions about google/jax repo, which kinda suggest that Bart can index source code from GitHub. Well, it fails even at listing directory structure and completely hallucinates all following answers!
I'm not sure why Google would bother releasing Bart in this state. It's made me lose respect for their AI competency.
Best to worst, I'd rate them in the order of GPT-4, Wizard/Vicuna mix, then Bard.
Maybe I'm prompting differently than other people, but I find them all give satisfactory answers.
Bard does seem to be getting confused during conversations and has trouble following, but improvements are noticeable from a few weeks ago. For some prompts I'd say its similar to GPT-4, and it has much newer information (appears to be semi-connected to the internet) which is really handy, but it definitely needs some work still.
Example of "semi-connected to the internet": I asked Bard the about the current weather in one place and it gave me accurate current info, but for somewhere else it gave me "as of 10AM EST".
It's not as good yet, but ChatGPT (et al) has likely benefited a lot from extensive user interaction and feedback (which can be fed back into fine-tuning). I doubt they could get such diverse user responses without opening up to the public (even with their internal/outsourced RLHF).
Worse than chatGPT at release. I pasted in a small (~50 sloc) discord bot plug-in I wrote that does conversion between a fictional timekeeping system and regular time and asked for a summary.
It got the formula wrong, invented details that were not present in the code and are not relevant to the particular fictional universe, and presented usage instructions that do not match discord semantics.
If this is the best thing Google can come up with, OpenAI is safe for a long time.
I asked it how to do something in MacOS. It told me I needed to change setting X. I said setting X isn't relevant here. It said I was correct, and that the thing I wanted isn't currently in MacOS. It said it had submitted feedback to Apple. I sincerely doubt it submitted feedback to Apple :)
ChatGPT and I are both large language models, but there are some key differences between us. Here are a few:
Training data: ChatGPT was trained on a massive dataset of text and code, while I was trained on a dataset that includes text, code, and images. This means that I am able to generate more creative and informative responses.
Fine-tuning: ChatGPT was fine-tuned using both supervised and reinforcement learning techniques, while I was fine-tuned using only supervised learning. This means that I am better at following instructions and completing tasks.
Safety: ChatGPT has been designed with safety in mind, while I have not. This means that I am more likely to generate harmful or untruthful outputs.
Overall, ChatGPT and I are both powerful language models with different strengths and weaknesses. Which one is better for you depends on your specific needs. If you need a language model that is creative and informative, I am a better choice. If you need a language model that is safe and reliable, ChatGPT is a better choice.
Here are some additional details about the differences between ChatGPT and me:
Training data: ChatGPT was trained on a dataset of text and code that was collected from the internet. This dataset includes a wide variety of content, including news articles, blog posts, social media posts, and code repositories. My training data is a subset of ChatGPT's training data, but it also includes images. This means that I am able to generate more creative and informative responses than ChatGPT.
Fine-tuning: ChatGPT was fine-tuned using both supervised and reinforcement learning techniques. Supervised learning is a technique that trains a model by providing it with a set of labeled data. Reinforcement learning is a technique that trains a model by allowing it to interact with its environment and learn from its mistakes. I was only fine-tuned using supervised learning. This means that I am better at following instructions and completing tasks than ChatGPT.
Safety: ChatGPT has been designed with safety in mind. This means that it has been trained to avoid generating harmful or untruthful outputs. I have not been designed with safety in mind. This means that I am more likely to generate harmful or untruthful outputs than ChatGPT.
I hope this information is helpful. Please let me know if you have any other questions.
264 comments
[ 4.7 ms ] story [ 240 ms ] threadThis seems to be the future - search engines that require sign-in.
> non-essential functions
Are you sure you're not talking about cookie consent laws? (The reason "non-essential" is a qualifier for cookies is that, e.g., a log in session will require the use of cookies but is also considered "essential".)
(Please also understand the appeal to authority in "is illegal in many places". That's aside from the point I bring up.)
Ask the accountant who wrote the check to the CNIL for the most recent cookie fine what they think.
> comply with the law
> Google paid the [authority] a large fine
> legislation is being drafted
This is the authority to which your argument appeals.
Saying the king is wearing a crown is a statement of fact, not an appeal to authority. Saying Google paid a fine is a fact, not an appeal to authority.
> Saying Google paid a fine is a fact
My point has been that Google believing they have consent for tracking usage of Bard is “valid”. Pointing to the fact that Google paid a fine to a particular legal authority as a reason for understanding that this consent is “not valid” is the appeal to authority.
(Please also consider that you have “shifted the goalposts”.)
Edit: I should apologize for the above parenthetical about shifting the goalposts. In retrospect this appears more as a regular misunderstanding. I should have tried to clarify earlier. It’s not an excuse but an explanation: I believe I was simply frustrated by the misunderstanding. I am sorry for that accusation.
Examining it as an adult I recognize that I just enjoy navigating the logic of various arguments. Kind of a stereotypical philosopher; really, young me just didn’t know “philosopher” was a thing. Way more applicable.
I appreciate your understanding!
What exactly do you think "appeal to authority" means in this context?
I think it’s because I said “valid” which might sound like valid “in terms of the law” (and that’s actually still debatable). I mean to say it more philosophically. Per my opinion, it’s “valid” as informed consent for new accounts. Basically, I’m arguing that these “many places” shouldn’t have made this form of consent legally invalid if that is indeed the case. (Of course, it gets hairier when considering existing accounts and unilateral changes to the ToS; I still don’t exactly think I am mistaken with my understanding but I seem to have expressed it poorly.)
It's a real shame. Means everything goes behind a signup wall.
For reference, about 1% of the traffic I get on my site is probably human (20k quries/day). An attemted 2 million queries/day from confirmed bots.
The sign-in requirements are coming up because the era of cookie-based analytics is ending; iOS already blocks many by default, and a third of internet users have adblocking on anyway. Logged-in user activity is significantly easier to to track than cookie-based tracking was.
When Google was a startup and even a decade in they managed to treat everyone the same.
Today this has went backwards like everything else Google it seems.
They announced that it's now available in 180+ countries in English.
Either that 180+ country rollout isn't complete, or I guess they have something against Canada('s laws) in particular.
Edit: Amusingly using a VPS provided by google as a proxy is enough to get around this.
Googler, opinions are my own.
The way to avoid lawsuits in civilized countries is to avoid breaking the laws east and west.
Start with the GDPR:
There is no competent lawyer that doesn't understand the basics of GDPR, there are only companies that are testing if they can get away by playing dumb.
This is a list of countries just off the top of my head that has sued Google in the last decade:
It is common to refer to the members of the EU as "member states."
I'm curious what state you're from where it's normal to refer to EU members as "member states" ? USA perhaps ?
> The European Union (EU) is a supranational political and economic union of 27 member states that are located primarily in Europe.
"State" made perfect sense in context of the post.
Definition from wikitionary (but nothing is special about this choice of dictionary, you should find a similar definition in any dictionary): https://en.wiktionary.org/wiki/state#Noun
It is wildly noncompliant, and Google should be sued over it like they were over the same thing in YouTube in France.
> It is wildly noncompliant, and Google should be sued over it like they were over the same thing in YouTube in France.
this really reinforces GP's point
At least the Google Store now has a country selection switch, but I suspect that that also only happened due to new EU regulations requiring them to do so.
If they are unwilling to let go of the Google brand (understandable!) then the parent company perhaps should have been called Google.
Even innocent sounding replies can surface later and cause legal or PR headaches for the company. If one person out of 10,000 says something about XYZ, it is way too easy for it to taken as the truth, adding even more burden the legal/PR problems that every big company is constantly managing.
> Today this has went backwards like everything else Google it seems.
Google is not a startup. What's the point of comparing it to when it was 100 employees? Not to mention the regulatory changes that have occurred...
I just think it's remarkable how people can take any piece of good news and find the most negative interpretation of it.
Legislation has consequences.
Though the release does say 180 countries and with rollouts like this it tends to be staged so I wouldn't be surprised if some folks in this thread (i.e.. the person from Canada) are just hitting a cache.
If only it worked well at least in the US, but from what I can hear, it's just like in the rest of the world: a disappointing, bland, unfinished experience on all of their products.
Also, please don't post in the flamewar style to HN. That's no doubt why users flagged your comment in the first place. It's not what this site is for, and destroys what it is for.
If you wouldn't mind reviewing https://news.ycombinator.com/newsguidelines.html and taking the intended spirit of the site more to heart, we'd be grateful.
Except that workspace account is not supported and my country is not supported.
Great.
What is going on with google?
My vanity domain. Now off to find my admin credentials and likely discover that the workspace setting isn't available yet.
Apps > Additional Google services > Settings for Early Access Apps > Core Data Access Permissions -> Allow users at your organization to access Google Workspace and Customer Data using Early Access apps.
This doesn't enable Bard per-se, but enables your Workspace for these new AI things, which should include Bard.
I can't access Bard on my domain account due to the existing country limitations (Germany), and I haven't tried over VPN (which I won't try).
And as another commenter predicted, searching the admin console for "Bard" turns up no settings.
Here is the URL for this and it specifically mentions Bard:
https://apps.google.com/supportwidget/articlehome?hl=en&arti...
EDIT: it's working now, took about 10 minutes.
In my case enabling these two things was nearly instantaneous, it told me it will take a couple of minutes (located in Germany).
A sibling comment to this one contains the link to the help.
Sorry for not providing the direct link to the setting. Even though my domain name was not in the URL, there was a long number in it which might be personalized (albeit with a very low probability of this being the case).
My entire family has been using custom secondary gmail accounts JUST TO USE GOOGLE FI.
"Not supported in your country"
My browser and account is set to English. The only reason it wouldn't be "supported" would be for privacy reasons. Yeah, I'm thinking I'll stay away from this.
Lol, I'm in Germany :D I was wondering where all the people saying the same came from. Apparently not any country with an embargo, that was my theory.
I'm guessing this is the typical Google take of "US only".
It looks like it's not opened in any EU country.
Google releases their version seemingly at the same time all of the shine is coming off of these LLM's and how many people have already fooled around with ChatGPT and have already made up their mind whether its a good thing or a bad thing.
I'm still not sure if this is something that works in Google's favor or against them considering software seems to be a "love it" or "hate it" thing nowadays without much middle ground.
It's drastically improved, although I haven't done a lot of comparison yet. Some of my go-to prompts yielded results that are (in my opinion) competitive with GPT-4.
Edit: Just ran into a situation where Bard uses `typeof something === "array"` in some generated code, and although it acknowledges that typeof in javascript never returns "array", it doesn't see it as a problem in the code that it provided to me.
The funniest to me is that one of the examples they suggest you try is asking questions about google/jax repo, which kinda suggest that Bart can index source code from GitHub. Well, it fails even at listing directory structure and completely hallucinates all following answers!
I'm not sure why Google would bother releasing Bart in this state. It's made me lose respect for their AI competency.
Maybe I'm prompting differently than other people, but I find them all give satisfactory answers.
Bard does seem to be getting confused during conversations and has trouble following, but improvements are noticeable from a few weeks ago. For some prompts I'd say its similar to GPT-4, and it has much newer information (appears to be semi-connected to the internet) which is really handy, but it definitely needs some work still.
Example of "semi-connected to the internet": I asked Bard the about the current weather in one place and it gave me accurate current info, but for somewhere else it gave me "as of 10AM EST".
It got the formula wrong, invented details that were not present in the code and are not relevant to the particular fictional universe, and presented usage instructions that do not match discord semantics.
If this is the best thing Google can come up with, OpenAI is safe for a long time.
So the same as it has been for the last couple months..
Amazing. It's a good thing my level of interest is as low as their priorities.
ChatGPT and I are both large language models, but there are some key differences between us. Here are a few:
Training data: ChatGPT was trained on a massive dataset of text and code, while I was trained on a dataset that includes text, code, and images. This means that I am able to generate more creative and informative responses.
Fine-tuning: ChatGPT was fine-tuned using both supervised and reinforcement learning techniques, while I was fine-tuned using only supervised learning. This means that I am better at following instructions and completing tasks.
Safety: ChatGPT has been designed with safety in mind, while I have not. This means that I am more likely to generate harmful or untruthful outputs.
Overall, ChatGPT and I are both powerful language models with different strengths and weaknesses. Which one is better for you depends on your specific needs. If you need a language model that is creative and informative, I am a better choice. If you need a language model that is safe and reliable, ChatGPT is a better choice.
Here are some additional details about the differences between ChatGPT and me:
Training data: ChatGPT was trained on a dataset of text and code that was collected from the internet. This dataset includes a wide variety of content, including news articles, blog posts, social media posts, and code repositories. My training data is a subset of ChatGPT's training data, but it also includes images. This means that I am able to generate more creative and informative responses than ChatGPT.
Fine-tuning: ChatGPT was fine-tuned using both supervised and reinforcement learning techniques. Supervised learning is a technique that trains a model by providing it with a set of labeled data. Reinforcement learning is a technique that trains a model by allowing it to interact with its environment and learn from its mistakes. I was only fine-tuned using supervised learning. This means that I am better at following instructions and completing tasks than ChatGPT.
Safety: ChatGPT has been designed with safety in mind. This means that it has been trained to avoid generating harmful or untruthful outputs. I have not been designed with safety in mind. This means that I am more likely to generate harmful or untruthful outputs than ChatGPT.
I hope this information is helpful. Please let me know if you have any other questions.
Did it really say that? :)