Brandon Paddock
@brandonlive.com
Architect for AI in Word at Microsoft.
Previous: ~14 years working on Windows, with a stint as small startup CTO in the middle. I also made Tweetium (may it rest in peace).
https://brandonpaddock.substack.com
Previous: ~14 years working on Windows, with a stint as small startup CTO in the middle. I also made Tweetium (may it rest in peace).
https://brandonpaddock.substack.com
You obviously did not see some of the posts I immediately reported and blocked. There were some really horrible ones.
December 21, 2025 at 8:38 AM
You obviously did not see some of the posts I immediately reported and blocked. There were some really horrible ones.
You didn’t see my mentions. Yes it was mostly trolls.
I don’t know where you’re getting that.
I don’t know where you’re getting that.
December 21, 2025 at 8:36 AM
You didn’t see my mentions. Yes it was mostly trolls.
I don’t know where you’re getting that.
I don’t know where you’re getting that.
I didn’t see that, but I responded now.
December 21, 2025 at 8:36 AM
I didn’t see that, but I responded now.
No, the point is that it depends on the kind of water being used and what the availability + competing usage is. And the overall usage for data centers is relatively very low, and usually of the least troublesome type. The point he was making was that the hyped concerns are largely overblown.
December 21, 2025 at 8:35 AM
No, the point is that it depends on the kind of water being used and what the availability + competing usage is. And the overall usage for data centers is relatively very low, and usually of the least troublesome type. The point he was making was that the hyped concerns are largely overblown.
So that’s one company and the issue there has been resolved? If other companies did similar things, then I think the same outcome sounds reasonable.
December 21, 2025 at 8:31 AM
So that’s one company and the issue there has been resolved? If other companies did similar things, then I think the same outcome sounds reasonable.
You don’t know me at all, and your arrogance is very off-putting.
December 21, 2025 at 8:29 AM
You don’t know me at all, and your arrogance is very off-putting.
I never said anything remotely like that.
December 21, 2025 at 8:27 AM
I never said anything remotely like that.
That’s exactly the kind of thing I work on!
I can’t control what investors do with their money or what people in the industry decide to hype. I think those ecological impacts are overstated (like the water thing), though I do think our energy policy with the current administration is problematic.
I can’t control what investors do with their money or what people in the industry decide to hype. I think those ecological impacts are overstated (like the water thing), though I do think our energy policy with the current administration is problematic.
December 21, 2025 at 8:27 AM
That’s exactly the kind of thing I work on!
I can’t control what investors do with their money or what people in the industry decide to hype. I think those ecological impacts are overstated (like the water thing), though I do think our energy policy with the current administration is problematic.
I can’t control what investors do with their money or what people in the industry decide to hype. I think those ecological impacts are overstated (like the water thing), though I do think our energy policy with the current administration is problematic.
What MIT study? One of what people think of or want from Microsoft’s AI offerings?
I’m doing my job, which is to make the product I work on better for our users.
I’m doing my job, which is to make the product I work on better for our users.
December 21, 2025 at 8:25 AM
What MIT study? One of what people think of or want from Microsoft’s AI offerings?
I’m doing my job, which is to make the product I work on better for our users.
I’m doing my job, which is to make the product I work on better for our users.
What a very bizarre interaction, and a strange threat.
I am proud of my work and the way I do it. I’m not perfect and I’m always learning, but I care and I help people, and I’m very good at what I do (and well respected for it). Your ignorant attempts to tear me down say a lot about you, not me.
I am proud of my work and the way I do it. I’m not perfect and I’m always learning, but I care and I help people, and I’m very good at what I do (and well respected for it). Your ignorant attempts to tear me down say a lot about you, not me.
December 21, 2025 at 8:17 AM
What a very bizarre interaction, and a strange threat.
I am proud of my work and the way I do it. I’m not perfect and I’m always learning, but I care and I help people, and I’m very good at what I do (and well respected for it). Your ignorant attempts to tear me down say a lot about you, not me.
I am proud of my work and the way I do it. I’m not perfect and I’m always learning, but I care and I help people, and I’m very good at what I do (and well respected for it). Your ignorant attempts to tear me down say a lot about you, not me.
I addressed both of those, neither of which supports the claims that *my work* (which I hope you understand is not at OpenAI or Google, never mind this other issues with those claims) is “destroying careers”.
December 21, 2025 at 8:12 AM
I addressed both of those, neither of which supports the claims that *my work* (which I hope you understand is not at OpenAI or Google, never mind this other issues with those claims) is “destroying careers”.
What “academic studies”?
December 21, 2025 at 8:10 AM
What “academic studies”?
Ah, more insults. You know what that says about your argument, right?
I don’t know what you even mean with that claim. But I am an expert on this subject matter and I know how these models work (including what they can and can’t do, and how they do it). What I said is correct.
I don’t know what you even mean with that claim. But I am an expert on this subject matter and I know how these models work (including what they can and can’t do, and how they do it). What I said is correct.
December 21, 2025 at 8:07 AM
Ah, more insults. You know what that says about your argument, right?
I don’t know what you even mean with that claim. But I am an expert on this subject matter and I know how these models work (including what they can and can’t do, and how they do it). What I said is correct.
I don’t know what you even mean with that claim. But I am an expert on this subject matter and I know how these models work (including what they can and can’t do, and how they do it). What I said is correct.
That is not an accurate reflection of the feedback we get from consumer and business users.
There is strong demand for more/better AI features, and our recent “agent mode” offerings in Office apps have been incredibly well-received by customers who have tried them.
There is strong demand for more/better AI features, and our recent “agent mode” offerings in Office apps have been incredibly well-received by customers who have tried them.
December 21, 2025 at 8:04 AM
That is not an accurate reflection of the feedback we get from consumer and business users.
There is strong demand for more/better AI features, and our recent “agent mode” offerings in Office apps have been incredibly well-received by customers who have tried them.
There is strong demand for more/better AI features, and our recent “agent mode” offerings in Office apps have been incredibly well-received by customers who have tried them.
You’re conflating different issues here (and I have no idea why you’d exempt DeepSeek given their training data is secret).
The model itself will not memorize a lone example from the training set. Oft-cited passages are more likely to be recitable with the right prompting.
The model itself will not memorize a lone example from the training set. Oft-cited passages are more likely to be recitable with the right prompting.
December 21, 2025 at 8:02 AM
You’re conflating different issues here (and I have no idea why you’d exempt DeepSeek given their training data is secret).
The model itself will not memorize a lone example from the training set. Oft-cited passages are more likely to be recitable with the right prompting.
The model itself will not memorize a lone example from the training set. Oft-cited passages are more likely to be recitable with the right prompting.
You say that like I’m being obstinate or something, but no one has provided any evidence to back up these claims.
(And “dragged” is an interesting way to put it. This all looks very different from my perspective.)
(And “dragged” is an interesting way to put it. This all looks very different from my perspective.)
December 21, 2025 at 7:54 AM
You say that like I’m being obstinate or something, but no one has provided any evidence to back up these claims.
(And “dragged” is an interesting way to put it. This all looks very different from my perspective.)
(And “dragged” is an interesting way to put it. This all looks very different from my perspective.)
Who is “they”? How do you know this?
So you’re saying “they” owe you the cost of purchasing one copy of your book(s)?
So you’re saying “they” owe you the cost of purchasing one copy of your book(s)?
December 21, 2025 at 7:52 AM
Who is “they”? How do you know this?
So you’re saying “they” owe you the cost of purchasing one copy of your book(s)?
So you’re saying “they” owe you the cost of purchasing one copy of your book(s)?
I am familiar with the accusation that training a model means the model is creating a copy or derivative work. That is the misconception I was addressing.
December 21, 2025 at 7:51 AM
I am familiar with the accusation that training a model means the model is creating a copy or derivative work. That is the misconception I was addressing.
I can assure you that is not true.
December 21, 2025 at 7:05 AM
I can assure you that is not true.
What? Where are you getting that idea?
I don’t know which companies paid to acquire what material or what other access they may have (rules around fair use for things like research and libraries can be complex), but I don’t think that’s what the concern was about.
I don’t know which companies paid to acquire what material or what other access they may have (rules around fair use for things like research and libraries can be complex), but I don’t think that’s what the concern was about.
December 21, 2025 at 6:47 AM
What? Where are you getting that idea?
I don’t know which companies paid to acquire what material or what other access they may have (rules around fair use for things like research and libraries can be complex), but I don’t think that’s what the concern was about.
I don’t know which companies paid to acquire what material or what other access they may have (rules around fair use for things like research and libraries can be complex), but I don’t think that’s what the concern was about.
But you didn’t address my response, which illustrates the problem with calling it “theft” to train a language model using publicly accessible and legitimately accessed works.
December 21, 2025 at 6:36 AM
But you didn’t address my response, which illustrates the problem with calling it “theft” to train a language model using publicly accessible and legitimately accessed works.
You’re going in circles. I already responded to that.
December 21, 2025 at 6:31 AM
You’re going in circles. I already responded to that.
Why wouldn’t you? It’s really one of the prime places for it. So much of the world’s work gets done there. Helping people get more done faster and easier is the core mission of productivity apps.
December 21, 2025 at 6:29 AM
Why wouldn’t you? It’s really one of the prime places for it. So much of the world’s work gets done there. Helping people get more done faster and easier is the core mission of productivity apps.
What a truly miserable, despicable person you are.
December 21, 2025 at 6:25 AM
What a truly miserable, despicable person you are.
What makes you think this is an acceptable way to talk to another person, let alone one whom you don’t know at all?
What value do you think you’re adding to the world by trolling strangers on the internet?
Go touch some grass.
What value do you think you’re adding to the world by trolling strangers on the internet?
Go touch some grass.
December 21, 2025 at 6:19 AM
What makes you think this is an acceptable way to talk to another person, let alone one whom you don’t know at all?
What value do you think you’re adding to the world by trolling strangers on the internet?
Go touch some grass.
What value do you think you’re adding to the world by trolling strangers on the internet?
Go touch some grass.