Inference is cash positive: it's research that takes up all the money. So, if you can get ahold of enough users, the volume eventually works in your favour.
This is VCs FOMOing as global-economy-threatening levels of leverage are being bet on an AI transformation that, by even the most optimistic estimates, cannot achieve a tiny portion of the required ROI in the required time.
IMHO, there's never been a better time to build your own product and learn to sell it. The effort that AI implementation requires is clearly exponential to complexity of the organization.
You can build faster now that you ever have: I am building faster than I have in 25 years of engineering. You have more capable support for all the unfamiliar processes of building a business imaginable.
And almost everyone larger than you is finding it harder to achieve similar productivity gains from implementing AI, if not outright struggling with it. This is a golden moment and won't last long.
I already get multiple cold emails a month selling me some sort of software product or service that I don't need and that wouldn't work for me.
If a bunch of additional people start going and building MVPs, what keeps that from becoming even more of a flood, like what you get if you post a job application on LinkedIn nowdays?
I think sales is likely to get harder, not easier, soon.
And to your earlier post, I think the big question is: can individuals get more done on their own now building new things because of less organizational issues compared to incumbents, or because of less product complexity and scope? Is the organization the problem, or is the complexity of stuff built to try to service a thousand different previous sales deals and customers?
Productivity has always been inversely exponentially correlated with manpower, I don't think this is related to AI, and I don't think AI speeds anything up in this scenario. If nothing else you're now constraining what could be one-man development speed to the speed of a chat-session, which is much, much slower.
If you know what you're doing (and most times even if you don't, yet) it's much faster to work actually solo, without bothering with any AI assistance except at most something like one-line auto-complete on command.
I can't agree with this more, and it's exactly both the position and the impetus of my personal action plan for the next several years to come.
Truly, AI levels the playing field, if you just have the inclination to look at it that way. It brings such potential to the right people. In my hands, a software developer who has been slinging code for food for 40+ years, I am a 'god' with a cheap monthly subscription to Claude Code. I can run several projects and keep track of every one of them and see progress like I've never had that chance to, in my entire lifetime. And yes, I still wield the stick, because I KNOW how to do this stuff. Claude really doesn't. Claude just gives me raw material, individual orchestral parts as it were, that I can put together with a conductor's baton.
My enterprise client? They get barebones slop. Enough to keep things alive but yet never be scalable and extendable. And they deserve that. They slaughtered our entire US based team and went overseas. They retain me to keep the wheels on while offshore wires up worthless, unpredictable AI agents to 'enhance' the product. I'm taking their money until I can break free sometime next year. They get the scraps now, and they will never know until karma brings them their bitter fruit.
My personal project. A really, really fun and fluid development momentum. Truly art and logic combined with Claude doing the grunt work for me. Absolutely a blast and it will have a user base, and the user base will really enjoy the results.
New partnerships. You can now make ANYONE you know, that much more viable, that much more tech savvy. As an individual, with your talents and our new little 'digital helpers', you, yourself can become the David to any Goliath you wish. You can help elevate ideas, small teams, and aspiring entrepreneurs and have a blast doing so.
Shed the notion that this moment has to be dark for you. It doesn't. It just takes seeing that maybe, just maybe, the boneheads that are trying to capitalize in these horrendous actions which include lying about AI, lying about the reasons they are laying off, etc. are bringing about their very demise in real time. Clueless as to what they are doing.
And those of us with the knowledge, ethics, high standards, discipline, and honor, are now armed with the very thing that can bring down those who created it to harm us.
I mean, _we_ probably can't think with our wetware on a read-only substrate. It doesn't establish it as essential, just that the only sure example in nature of thought doesn't work that way.
I'm not an expert but as far I understand, plasticity is central to most complex operations of the brain and is likely to be involved in anything more complex than instinctive reactions. I'm happy to be corrected but it is my understanding that if you're thinking for a while on the same problem and establishing chains of reasoning, you are creating new connections and to me that means it's fundamental in the process of thinking.
Also not an expert :) I thought plasticity is an O(hours-days) learning mechanism. But I did some research and there is also Short Term Plasticity O(second) [1] which is a crucial part of working memory. We'd need that. But it seems it’s more of a volatile memory system, eg calcium ion depletion/saturation at the synapse, rather than a permanent wiring/potentiation change (please someone correct me if this isn’t right :) ).
So I guess I’d just clarify “read only” to be a little more specific - I think you could run multiple experiments where you vary the line of what’s modeled in volatile memory at runtime, and what’s immutable. I buy that you need to model STP for thought, but also suspect at this timescale you can keep everything slower immutable and keep the second-scale processes like thought working.
My original point still stands - your subjective experience in this scenario would be thought without long-term memory.
It would be funny if what you get from a read only human brain is a sort of memento guy who has no capacity to remember anything or follow a conversation... kind of like an LLM!
Google's own Gemma models are runnable locally on a Pixel 9 Max so some lev of AI is replicatable client side. As far as Gmail running locally, it wouldn't be impossible for Gmail to be locally hosted and hit a local cache which syncs with a server only periodically over IMAP/JMAP/whatever if Google actually wanted to do it.
The gain, as far as local AI goes for Google, is that, at Google scale, the CPU/GPU time to run even a small model like Gemma will add up across Gmail's millions of users. If clients have the hardware for it (which Pixel 9's have) it means Gmail's servers aren't burning CPU/GPU time on it.
As far as how Gmail's existing offline mode works, I don't know.
To make the discussion easier, we can look at the state of GMail before widespread AI was a thing.
(I have to use some weaselwording here, because GMail had decent spam detection since basically forever, and whether you call that AI or not depends on where we have shifted the goalposts at the moment.)
Heh that brings to mind that old chestnut, something like "We call it AI until it works, then we call it machine learning." Google's running the servers tho, so the assumption is the servers would be running the spam filter against it before the user. In my mind, anyway. So it's fair to point out that Google also has the compute resources to not have to care. To the hypothetical though, could Google make Gmail a local-first? I say yes. Will they? I doubt it.
