I’m having trouble wording this properly: Even though I am happy there is a familiar name taking on the task of ethical AI after Google and Microsoft made their stance clear by firing their own ethical AI teams, Mozilla should probably focus on their existing portfolio’s stability before picking up a project of this magnitude.
I agree. I’m not unhappy with the vision, but I’m growing ever so increasingly worried about Mozilla’s inability to stabilize their core ventures instead of taking weird side projects that they drop faster than Google does.
Mozilla Corp and Foundation had access to talent, goodwill and community to be the leaders of a true alternative to FAANG.
Instead, the world’s best browser is stagnant; infused with a featureless password manager, a resold third party VPN and a ‘me too’ email relay plugin.
The Foundation is busy attending to “trustworthy” AI, “re-imagining” web monetisation thru tech (a legislation issue) and trying to make their unaccountable competitors ethical. “New approaches to data stewardship,” an absolute surrender to the violent status quo of the serialisation of everything.
$30M is a tiny amount. OpenAI has over $10B in funding. It’s hard to get precise numbers, but it sounds as if Deep Mind is spending over $1B/year on compute now. It’s unclear how Mozilla expects to play in the same playground with so much less money.
And they don’t have any product to stick it on. OpenAI offers it as a product, and Microsoft/Facebook/Google want it for search engines and ads. Who is Mozilla helping here, if not the manufactured Silicon Valley hype?
[ Disclaimer: I work for Microsoft, but not on AI things ]
I don’t think ads are a big driver for our management. Integration with things like Office and selling services like Copilot are (they seem to want copilot-like things everywhere. I want a better Dall-E in PowerPoint for illustrating slides). The big revenue, I expect, comes from the hype. The OpenAI deal means that Azure is selling the only hosted GPT-4 offerings for people to integrate in other products, so everyone that buys the ‘GPT-4 [French for ‘I farted 4’, which sounds like a terrible film] will change everything’ hype becomes an Azure customer.
Oh yes you are right. Microsoft stands out from the other two by offering many enterprise products. (I’m not aware of any plans of Google integrating Bard with their Workspace product)
Agreed. The TPUs required for training anything of competitive value will demolish $30m. The sort of researchers you’ll need probably command a $500k salary. $30m is basically hiring and office space.
Im confused at the negative comments, it feels like this is exactly what a big non profit tech company should focus on considering the insane amount of AI misuse right now.
Mozilla is more than just for tech nerds, normies will use AI and Im not aware of any ethical AI organization besides potentially this.
In part, I think the reactions you’re seeing are because people don’t feel Mozilla has been succeeding particularly at its core mission.
Throwing $30 million at a problem that the other tech giants are throwing hundreds of millions of dollars ad seems like tilting at a windmill. This comes just a few years after Mozilla laying off quite a few people who worked on its core products. But, sure, let’s try to jump on the AI bandwagon.
I just don’t feel like Mozilla’s current leadership is going in the right direction generally. They’ve presided over a continually shrinking market share for Firefox and don’t really seem to have any ideas for turning that around to justify their inflated salaries. So now they’re going to sprinkle some AI on it to look like they’re doing something. Spiffy.
Would it be nice if this succeeds? I guess. Am I optimistic that it will turn into anything real? No.
Off the top of my head, other ethical AI communities include DAIR for discussion and Hugging Face for sharing datasets and models. Reading through Mozilla’s whitepaper, Creating Trustworthy AI, I am not sensing a concrete proposal. At best, it sounds like they may copy Hugging Face somewhat, reworking some of their existing machine-learning tools for a wider audience.
That said, I would love to see Mozilla exceed expectations.
Mozilla came from the ashes of Netscape Communications Corp to give us a competitive open-source browser. Anything outside of that is scope creep. This is well outside of that. We don’t need them failing at two things.
There isn’t a “big non-profit tech company”…there’s the Mozilla Foundation, and the Mozilla Company, and basically a continued de facto existence as anti-trust insurance for Google.
There is not obviously an “insane amount of AI misuse”. This is a bold claim–potentially true, but citations needed.
“Ethical AI” is incredibly vague. Is an AI ethical that reliably answers the questions you pose it? Is an AI ethical that follows Asimov’s 3 laws? Is AI ethical that refuses to account for some political or social good in its operation? What if that good is subjective? Is an ethical AI one that reflects the values of its owner or operator?
Is this ethical AI research just going to be paying a bunch of Yudkowsky fanboys and AI skeptics to jerk each other off in thought experiments and papers, when the bug tracker for Firefox is alive and well?
Is 30MUSD even going to make a dent on GPU training costs for models?
Given Mozilla’s own history, do we actually expect them to be unbiased purveyors of AI (cough Eich’s railroading cough)? Should they be?
There are plenty of reasons many of us are bearish on this move.
I’d like to prefix this comment by saying I’m a happy and loyal Firefox user, and have been for many years. My laptops and phones all run Firefox; I use it on (at last count) FreeBSD, Manjaro, Ubuntu, and Mint. With that said …
As a consultant I’ve worked on a few projects where the stakeholders were afraid to try to win.
For example, one outfit hired me (and a team of astoundingly skilled developers - some valuable open source tooling and at least one spin-off startup came out of that project) to fix a problem with their back-office systems. The underlying problem was that everything was bolted into an obsolete, heavily customised, non-upgradeable, non-tested, legacy CRM installation.
It became apparent that management would do anything but let us attack the actual mission. We were encouraged to mess around the edges, working on “MVP” solutions to imagined or low-value problems, while the cancer that was the CRM sat untouched. Any suggestion of replacing the CRM, or even putting the brakes on work that would increase coupling to it, was career poison.
Looking back on that with the benefit of years of hindsight, I think it was driven by incentives. You could take a swing at the actual problem, with a reasonable chance of failure. Or, you could attempt something much easier, with a higher chance of success, and get promoted / receive a bonus / leave for a higher paying job. All the while taking tours of enterprise sightseers for walks around the company to see how “Agile” we were.
I wonder if there’s a similar thing going on here with Mozilla? Their mission - although I know they would disagree - is to keep the open Web alive by making Firefox the best it can be. Provide a genuinely open source, privacy respecting, feature-rich, high-performance alternative to Chrom* and Safari. One that is capable of enticing people away from the other browsers by being better.
But that’s a hard mission. It always was; but the mainstream browsers nowadays are much tougher competition than IE6 was, and they’re even more entrenched. It might even be a doomed mission today; perhaps the failure now is in refusing to admit that?
But please, Mozilla. Please, please, please, try to focus on the key mission instead of fluffing around at the edges.
I’m having trouble wording this properly: Even though I am happy there is a familiar name taking on the task of ethical AI after Google and Microsoft made their stance clear by firing their own ethical AI teams, Mozilla should probably focus on their existing portfolio’s stability before picking up a project of this magnitude.
Unfortunately, this seems to be very well worded 😭
I agree. I’m not unhappy with the vision, but I’m growing ever so increasingly worried about Mozilla’s inability to stabilize their core ventures instead of taking weird side projects that they drop faster than Google does.
100%. If you’re in search of more words, I just remembered this thread by Cade Diehm from a few months ago: https://post.lurk.org/@shibacomputer/109548784095427843
I don’t know, this seems a bit hyperbolic. I’m perfectly happy on Firefox, the VPN worked well when I used it, and I love relay.
Well I’m happy I stopped donating to Mozilla years ago:
$30M is a tiny amount. OpenAI has over $10B in funding. It’s hard to get precise numbers, but it sounds as if Deep Mind is spending over $1B/year on compute now. It’s unclear how Mozilla expects to play in the same playground with so much less money.
And they don’t have any product to stick it on. OpenAI offers it as a product, and Microsoft/Facebook/Google want it for search engines and ads. Who is Mozilla helping here, if not the manufactured Silicon Valley hype?
[ Disclaimer: I work for Microsoft, but not on AI things ]
I don’t think ads are a big driver for our management. Integration with things like Office and selling services like Copilot are (they seem to want copilot-like things everywhere. I want a better Dall-E in PowerPoint for illustrating slides). The big revenue, I expect, comes from the hype. The OpenAI deal means that Azure is selling the only hosted GPT-4 offerings for people to integrate in other products, so everyone that buys the ‘GPT-4 [French for ‘I farted 4’, which sounds like a terrible film] will change everything’ hype becomes an Azure customer.
Oh yes you are right. Microsoft stands out from the other two by offering many enterprise products. (I’m not aware of any plans of Google integrating Bard with their Workspace product)
Agreed. The TPUs required for training anything of competitive value will demolish $30m. The sort of researchers you’ll need probably command a $500k salary. $30m is basically hiring and office space.
Im confused at the negative comments, it feels like this is exactly what a big non profit tech company should focus on considering the insane amount of AI misuse right now.
Mozilla is more than just for tech nerds, normies will use AI and Im not aware of any ethical AI organization besides potentially this.
We should want this to succeed.
In part, I think the reactions you’re seeing are because people don’t feel Mozilla has been succeeding particularly at its core mission.
Throwing $30 million at a problem that the other tech giants are throwing hundreds of millions of dollars ad seems like tilting at a windmill. This comes just a few years after Mozilla laying off quite a few people who worked on its core products. But, sure, let’s try to jump on the AI bandwagon.
I just don’t feel like Mozilla’s current leadership is going in the right direction generally. They’ve presided over a continually shrinking market share for Firefox and don’t really seem to have any ideas for turning that around to justify their inflated salaries. So now they’re going to sprinkle some AI on it to look like they’re doing something. Spiffy.
Would it be nice if this succeeds? I guess. Am I optimistic that it will turn into anything real? No.
Off the top of my head, other ethical AI communities include DAIR for discussion and Hugging Face for sharing datasets and models. Reading through Mozilla’s whitepaper, Creating Trustworthy AI, I am not sensing a concrete proposal. At best, it sounds like they may copy Hugging Face somewhat, reworking some of their existing machine-learning tools for a wider audience.
That said, I would love to see Mozilla exceed expectations.
Some thoughts:
There are plenty of reasons many of us are bearish on this move.
I’d like to prefix this comment by saying I’m a happy and loyal Firefox user, and have been for many years. My laptops and phones all run Firefox; I use it on (at last count) FreeBSD, Manjaro, Ubuntu, and Mint. With that said …
As a consultant I’ve worked on a few projects where the stakeholders were afraid to try to win.
For example, one outfit hired me (and a team of astoundingly skilled developers - some valuable open source tooling and at least one spin-off startup came out of that project) to fix a problem with their back-office systems. The underlying problem was that everything was bolted into an obsolete, heavily customised, non-upgradeable, non-tested, legacy CRM installation.
It became apparent that management would do anything but let us attack the actual mission. We were encouraged to mess around the edges, working on “MVP” solutions to imagined or low-value problems, while the cancer that was the CRM sat untouched. Any suggestion of replacing the CRM, or even putting the brakes on work that would increase coupling to it, was career poison.
Looking back on that with the benefit of years of hindsight, I think it was driven by incentives. You could take a swing at the actual problem, with a reasonable chance of failure. Or, you could attempt something much easier, with a higher chance of success, and get promoted / receive a bonus / leave for a higher paying job. All the while taking tours of enterprise sightseers for walks around the company to see how “Agile” we were.
I wonder if there’s a similar thing going on here with Mozilla? Their mission - although I know they would disagree - is to keep the open Web alive by making Firefox the best it can be. Provide a genuinely open source, privacy respecting, feature-rich, high-performance alternative to Chrom* and Safari. One that is capable of enticing people away from the other browsers by being better.
But that’s a hard mission. It always was; but the mainstream browsers nowadays are much tougher competition than IE6 was, and they’re even more entrenched. It might even be a doomed mission today; perhaps the failure now is in refusing to admit that?
But please, Mozilla. Please, please, please, try to focus on the key mission instead of fluffing around at the edges.
Did any of these Mozilla ethics projects turn into something?