No it's the frog boil effect. It's a staggering amount of change, it's just over a whole human lifetime so most people don't really clock it. Elites are supposed to do something about it, but mostly because of corruption and ideology they aren't.
6,141 karma · joined November 8, 2014
And for some perspective, this is only one of many other huge changes that huge populations will react violently to in the next 20-50 years. Good luck to us all.
> The reason her startup was growing so fast was simply that users loved what she'd built.
> In the real world, growth rates tend to slow down a bit. A very successful startup will probably be growing faster than 15% a month in year 1 and slower than 15% a month in year 4.
It turns out that the people who will invest in your startup when 93% MoM gains are possible want you to do pretty much anything to keep growth as high as possible--also your career, net worth, and employment are tied to this so you're similarly motivated--including squeezing and manipulating those users who loved you so much. But hey, as long as you personally get rich it's fine I guess.
Another way you could take it is, avg Uber salary is what $300k/yr? Does Uber think LLMs make their engineers at most .5% more productive?
- Take a selfie
- Create a reminder
- Call Vicki
- Rotate photo left
- Create a new event (do you create an old event?)
- Send an email
- Resume my podcast
- Create a note
- Add photo to album
...can't iOS do literally every one of those things already? What the fuck is happening?
Very true, mostly because it's pricey to do so--or more specifically, putting a server in DCA is $.005/user, putting a server in JNB is like $.50/user sooooo....
Sometimes this is a "so what", other times it has huge implications for app architecture.
This doesn't look at all exponential to me: https://epoch.ai/benchmarks?view=graph&tab=eci. OpenAI models went from 137 ECI to 159 ECI over about a year and a half, and the trends are similar for Anthropic and Google. These things have never been exponential.
> The models optimise for solving the benchmark and we create new benchmarks to assess broader intelligence. As models converge on 100%, progress obviously slows.
We are nowhere near 100% on important benchmarks like hallucinations: https://artificialanalysis.ai/evaluations/omniscience?model-...
...also, progress isn't improving with model releases.
---
We're running out of money. While we don't know how much it cost to train things like Claude, most (all?) industry reports indicate that a significant gain in function (2x) would require an exponential amount of resources (20x). No one's yet been able to convince investors that's worth it.
Also, we're running out of data: https://epoch.ai/publications/will-we-run-out-of-data-limits....
Also, we're running out of of low hanging fruit: "We find that the level of compute needed to achieve a given level of performance has halved roughly every 8 months, with a 95% confidence interval of 5 to 14 months. This represents extremely rapid progress, outpacing algorithmic progress in many other fields of computing and the 2-year doubling time of Moore’s Law that characterizes improvements in computing hardware (see Figure 2)." (https://epoch.ai/publications/algorithmic-progress-in-langua...). Maybe you think we'll continue along this breakneck pace, but again no investor thinks that, which is why prices are going up (investment is drying up).
Also we're running out of compute. Data center projects are stalling. Some of this is spiking energy prices, some of this is politics, much of this is grid constraints and supply chain problems: https://tech-insider.org/us-ai-data-center-delays-cancellati....
---
Finally, and perhaps worst of all, despite unprecedented investment data on the productivity gains is mixed. This is the biggest difference from other technological leaps like electricity, the industrial revolution, literally fire, etc. Those things were immediately, undeniably more productive. AI is not like that. You're not seeing an AI Microsoft, an AI Salesforce, an AI Oracle, an AI SAP, etc. You can argue that their advantages are structural, but there are no successful AI-powered alternative products (no AI Office, no AI ERP, no AI database, etc).
- Mandatory screening of students in grades K-3, three times a year, using approved assessment tools, to monitor how early reading skills are developing
- Focused efforts to improve curriculum quality in schools— for phonics and other aspects of literacy
- Large-scale efforts to train teachers
- Retention policies to hold back students who aren’t reading successfully by the end of third grade
Lifted directly from https://www.theargumentmag.com/p/is-mississippi-cooking-the-...
It's also worth saying a lot of the gains evaporate when you factor in learning loss from COVID and Mississippi suddenly screening/helping kids for dyslexia (etc). But, still there's something to learn, just not simply phonics.
Two things I would say. One, reliability is real bad. Sometimes getting devices connected is a ridiculous pain in the ass. Sometimes automations just don't work. Sometimes devices just go dark and can't be brought back. Etc. etc. etc. So, you don't really want to engage, and you don't really want to make it a critical part of your home. Or said another way: it prevents you from tinkering in a fun way. When kids were learning BASIC on their home computers, things were pretty reliable. Not always, I mean hardware failure or whatever happens. But it was good and the feedback cycles were short. You're not gonna get people curious and thus experts with the existing smart home devex.
The second is we expect practically nothing from people in our culture. You don't need to be an expert in radio communications to know like, more walls between a device and your AP is bad, or that channels can get crowded. You don't need to be a tech wizard to know how to reset a device (you look in the manual or search online). We should be embarrassed that some of the most educated and wealthy people in history can't perform even basic logic or troubleshooting.
You can tell a human (an IC) the same things. You can then tell another human (a manager) "hey, check that IC A did these things". So far, these are the same. But there's now a critical difference: you can then hold those people accountable if they don't. They can be in the critical path. LLMs can't. People can improve. LLMs can't. People can work together. LLMs can't.
This doesn't always matter. You don't need things like accountability or improvement or teamwork all the time. But you do in reliable software.
What do you think is more likely: cops fudging facts on reports super frequently because it works, or PDs and their shared investigator having time to run down every report for each of their clients and courts accepting the findings?
These things just can't be in the critical path. They are ridiculously unreliable.
Or like, in your example (to be clear I defer to you entirely on the facts), not only would Foxx have to be a competent State's Attorney, she'd have to be competent at changing a system that's not ideal for her into one that is, e.g. by persuading existing staff, hiring new staff, etc.
This is by no means impossible--it's called leadership. My hot take, more or less, is that if you get someone into office who shares your values but can't make the office execute those values, not only will you fail but you'll tarnish your political movement for... a good long time, so competence is actually more important than values alignment.
My even hotter take is that people used to know this. When people from minority groups or political persuasions got power (women, LGBTQ people, racial/ethnic/religious minorities, ideological/philosophical minorities, etc) they knew they were modeling for everyone in their group. Is it fair? No. Is it reality. Super yes! Is whining about unfairness good politics and/or attractive at all? Super no!
Also if the threat model you're addressing w/ VPN usage is anything other than "I don't want my ISP to know what I'm doing" you need to use/do something else.
For us (my partner and I) these discussions dovetail into discussions about community. Like, so much about modern, suburban, nuclear family stuff is really isolating for everyone involved. We don't know exactly where to go from here, but looking at the declining fertility rate, it does seem unsustainable.