I hope you post this again when you do - I was presented with the "0.00080" difference a couple times, and it looks like this is where it becomes actually impossible because of this issue.
It’s cute, and I’m trusting enough to believe them when it says 100% home made, but square images with a strong yellow tint will forever be associated with ChatGPT 4o image generation in my mind.
Unfortunately, this might become something like the em-dash—where artists start tweaking their work to look less like the AI’s that are copying them.
> Unfortunately, this might become something like the em-dash—where artists start tweaking their work to look less like the AI’s that are copying them.
So true! (And yes—I see what you did there.)
It's even happening to photos now. A few months ago I posted a "Bot alert!" on Nextdoor warning people about the latest scambot.
One person replied "It's funny to see a bot reporting a bot."
I asked how they discovered I was a bot.
"It's your profile photo. The facial expression is too good, and the smoothness of the background is too perfect. Has to be AI."
For the curious, it's the same photo as on my LinkedIn:
What they didn't know was how I took that selfie. I set up my Micro Four Thirds camera on a tripod in the front yard, with the world's best portrait lens: the Olympus 75mm f/1.8. I stood some 10-15 feet from it (this lens is equivalent to a 150mm lens on a full frame camera, i.e. a moderate telephoto) and used the remote control to take a few dozen shots as I let my face relax into various expressions.
I picked out 4-5 favorites and asked a friend about them. She said "This one. It has gravitas."
I don't even think it's that great a photo. But I suppose the "gravitas" makes it look like AI.
For a photo that really shows off what that 75mm lens can do, check out this one of our late dog Brownie, titled Pumpkin Brownie:
The cheese pattern and the green teacup pattern after it are obviously AI generated. The weird curve of the wedges, the fuzzy edges to the cheese holes, the artifacting around the edges of the teacups, the fact that neither is a perfectly repeating pattern. It's 100% AI, even if the font may not be.
100% Homemade is just a stock phrase that they are using to display the type-face. I don't think you should take that to mean anything more than "Feathers McGraw."
LLM cliches are just condensed real world cliches.
Work as middle of the road as this sits right at the heart of that. It’s supposed to be warm and it’s entirely digital, hence the ways of conveying warmth are the same.
I have worked with Aardman. Unsurprisingly everything is shot digitally.
The gyro permission request doesn't work on iOS since it's not tied to user input.
If you're feeling brave, you can paste this into your phone's javascript console to add a button that requests permission.
var b=document.createElement('button');
b.textContent='Gyro';
b.style='position:fixed;z-index:999';
b.onclick=()=>{DeviceOrientationEvent.requestPermission();b.remove()};
document.body.appendChild(b);
It’s different if you have influence over the network, like a government might. I spend a lot of time in China, and they’ve done a good job of blocking VPNs in recent years, including my personal WireGuard connection to my home network. Not that any technical solution is impossible to bypass, but a motivated state government could make VPN use difficult if it wasn’t for the whole Constitution thing.
Pretty lacking in math for an article that mentions math a few times.
I looked into this a few years ago when I was trying to see if we were really in the worst housing market ever, and came to the opposite conclusion.
https://arriens.us/articles/housing.html
I used this method to check some of the lights in my house a few years ago.
The slow-mo video mode on my phone used a rolling shutter which captures one row of pixels at a time, meaning you could see the flicker in part of the video, even when it’s a multiple of 60 Hz or above 240 Hz. The flicker and camera frequencies also aren’t exactly synced up, so you can see the dimmed parts move across the screen.
You can get a pretty good idea of frequency, depth of flicker, and if the LED’s colors are flickering in sync from this, and I can confirm that Philips LEDs, specifically the EyeComfort series, are good.
I browse Hacker News through a custom aggregator. This post is how I found out it’s susceptible to HTML injection - a (2020) was marqueeing across my screen.
"All failure is shear failure" - this is a simple explanation of Tresca's Yield Criterion. For materials with higher compressive than tensile strength, the equivalent is the Mohr-Coulomb failure criterion.
For a simple example try twisting a ductile aluminum bolt (or clay), twisting a brittle piece of chalk (or concrete), or twisting a composite (twig). They all fail differently (and the first two at 45deg to each other. Mohr's circle is interesting, and fatigue failure more interesting still.
Before having children, I was quite interested in the idea of, and the math behind early retirement.
Most of the interesting math happened at the margin: you’ve got just enough money that you could retire, but you’re susceptible to risk of a market crash in the first few years of retirement or an abnormally long life expectancy combined with a middling market.
Tontines fascinated me as an interesting piece of the puzzle for those who don’t plan on leaving an inheritance, and I’ve reread this guide[0] a few times - but ultimately it’s just another way to possibly move the margin a little bit, and the real solution is to save a little bit more, then spend a little bit less.
Ahead of the launch of the first Tontine company in over 100 years, we have heard similar comments from many parents and for that reason have now created the Tontine Trust Fund.
The regular Tontine Trust is for parents that want to avoid the risking of running out of money in old age and becoming a financial burden on their children.
The Tontine Trust Fund is for parents that want to set aside an inheritance for their spouse or children now which they can configure to start paying the child a monthly income for the rest of their life starting at age X. This reduces the concern of parents that they will pass on a chunk of the inheritance to children that will 'blow the money' instead of making it last them for life.
Also, FYI:
a) Research from the insurance industry indicates that tontiners/annuitants spend double what they would without having a lifetime income, thereby enabling a better quality of life in retirement.
b) The Swiss Federal Institute of Technology, alma mater of Einstein and 28 other Nobel Prize Winners, has produced research showing that a retirees pension wealth is enhanced by 87% with zero added risk upon moving their savings into a Tontine, indicating that the gain is not 'marginal'.
All in all, the Tontine enables you to save a little less yet still spend more.
Those rates and risks are meaningless without a baseline, as Einstein and 28 other Nobel Prize Winners may agree.
If you're familiar with the early retirement community, the simplest strategy is withdrawing a fixed percentage of your initial retirement portfolio, adjusting for inflation every year. For an 100% equities portfolio, these are the odds of success over a 30 or 60 year horizon[0] when backtested against Shiller's total real return data from 1871-2018
4%/30 year: 97%
4%/60 year: 89%
3%/30 year: 100%
3%/60 year: 100%
Hence my comment about spending a little less or saving more - 4% to 3% makes a massive difference in success rates. I'm sure you've done some backtesting of your offerings, and hopefully would be able to share some withdrawal amount vs success rate comparison, even if it's not an identical time period/comparison.
I mean, this just appears to be fixed nominal returns, minus a fee, multiplied by a factor from an actuarial table.
Contrast it with a calculator like this [0] that uses combines historical return and inflation data with actuarial data to show the variance of outcomes, not just average returns.
For instance, your calculator shows a scenario of investing in bitcoin and withdrawing >20% of your portfolio every year which makes zero sense once you account for variance of returns.
I like the idea of tontines, I'm glad someone is trying to bring them back, and I don't doubt that your product could help with longevity risk, but I haven't seen anything so far that actually shows that.
I'd like to see actual results from backtesting, or a prediction that takes risk into account, not just a fixed return.
