I used to work in WebRTC back in it's earlier days and our team developed the open-source rtc.io. (https://github.com/rtc-io)
I never would have imagined that OpenAI is sending the full audio of a request to their servers. I had always assumed the audio was transcribed locally and then sent to the server.
The only reason I can think they'd want the full audio is for later model training, which, ok, fair-enough, but this can still likely be done without the limitations of WebRTC.
I can't think of a single time I've looked at a product tour and thought "well, I'm really glad they told me that, I never would have figured that out.
What the product tour I think often misses is that people don't want to learn your entire tool at one time.
They came to do one thing, that one thing needs to be brain dead simple.
Over time, you can show people what else they can do. But a product tour isn't the way to do that.
I think progressive UIs where you expose more and more to the user over time is the way to go.
If you're thinking "but I have so many features and capabilities this person needs" you probably haven't identified what the one thing people are paying you for is.
Computers were nowhere for ever, then everyone had them.
The internet was tiny, then everywhere.
Smartphones were a teensy market, then everyone had them.
GLP1s were for a small group of diabetics, now a significant portion of the population take them.
This is how things playout time and time again.
Does it mean the commentors 10 years is correct? No. But it also doesn't need to be incredibly optimistic. All it takes is getting the robots right, and there are multiple companies who seem very close.
Mostly just the cost, yeah. It will be like buying a car. The economics will have to make sense for regular people, while it starts popping up in tons of places and become a status symbol.
Digital computers existed for ~10-20 years before hitting the consumer market. It took almost a half-century for the microprocessor to become a ubiquitous appliance.
Interesting comment they have towards the end about "targeted memory reactivation can disrupt sleep".
It is important to note the study they are referring to is "targeted memory reactivation with sleep disruption", there are methods of doing targeted memory reactivation without sleep disruption.
I work in neurotech/sleeptech as the founder of affectablesleep.com, and though we are mostly focused on slow-wave (deep) sleep, we have been looking into memory reactivation, lucid dreaming and other stimulations for additions.
Sorry for the off-topic, but I was curious about Affectable so I opened the website. I saw it's very thin and light and comfortable, but I struggled to find out what "it" is and what it does for me. It's kinda buried.
I was interested enough to click through the different links in the footer. And just as I reached the purchase page, I see that it requires "an iPhone running iOS". Unsure why it requires an iPhone; and no info on a timeline for iPhone-less customers. But that immediately rules me out as a customer.
I feel like the landing page would be a lot better if it started out focusing on what it is & how it can help me.
Apologies again for the unsolicited advice. Just wanted to share my impressions in case it's helpful.
I used to do this regularly when I first started coding, I called them "Codemares". They were like nightmares with the shouting of commands I didn't quite understand would invade my dreams.
It seems to me that this is the purpose of nightmares. I especially noticed this after having kids. They are not by default scared of snakes and such but if they see a nature documentary of a snake biting something or even a cartoon bad guy, it's enough to trigger bad dreams which reinforce the fear and it's far stronger the next day.
IMO this is under-appreciated in current AI models. RL is not very effective in avoiding crocodiles for example, by the time like 5 of your tribe-mates are eaten it's far too late. You need some mechanism that ensures the danger is learned after just a single incident.
According to google, this would be almost 30% of total US energy production (135gw-150gw) and nearly 5% of total US energy consumption.
But what is the "breakthrough" if there is one? The article doesn't really suggest any breakthrough that is unlocking this potential energy? Or maybe I'm looking for a technological breakthrough where there isn't one.
There isn't one. They are trying to politically pressure a utility to build some geothermal plant. But utilities have engineers who will tell their bosses that this plan doesn't work. So the companies selling the geothermal plant are trying to politically pressure the utility to do yet another thing that they know won't work. PG&E for example has several geothermal plants which have been economic disasters and were and are being shutdown.
The core breakthroughs were working with partners to develop PDC bits that enable high rates of penetration in drilling out these horizontal wells in high temp granitic rock and then demonstrating plug / perf fracture networks that have a high engineered permeability in these source rocks to support economical flow rates and heat transfer. These were considerable advances over previous efforts.
There will be other learning by doing advances in how you structure your power plant design to take advantage of these to make practical long term power production possible (well spacing and injection / production placement / flow rate and temperature decline management).
They're adapting fracking techniques to use for geothermal, which opens up many more sites. Historically geothermal has had limited potential, and the best sites have already been developed. So geothermal + fracking creates a lot more viable land.
Traditional geothermal is you dig a really deep well and get a geyser of hot water or steam to come up.
Fervo is doing "Enhanced" or "Engineered" geothermal where you dig two wells: an injection and an extraction well. You frack the rock in between, creating lots of small channels for water to flow between them. The water absorbs the heat from the rocks as its circulating from the injection well to the extraction well.
The kind of rock that's good at this heat transfer is different from shale rock that oil & gas frackers have experience with... it's harder, less porous, not partial-dino-juice. So they're taking a lot of the same core concepts from the oil & gas industry (horizontal drilling, geology simulations, etc), but their IP is in adopting the techniques to work with geothermal-favorable rock.
Another interesting concept I heard Fervo researching: this kind of geothermal is not "baseload" style power, so there's a few tricks they can do to get better cost efficiency and peaker-like or battery-like behavior. Remember the two wells that form the circulation loop: injection and extraction? Well, you need pumps on both sides (remember, this isn't "geyser-style" geothermal where natural pressure and geology do all the work). Pumps take energy to run, something like 20-30% of the overall extraction output (you put a unit of energy in to run the pumps and you get 3-5 units of energy back out the other end). Not great, not terrible either... it's an energy return comparable to solar and wind. But what you can do is run the injection pump when power prices are low (ie when there's an excess of solar on the grid), pressurize all your fracked channels underground (the reservoir), and then when grid prices rise in the evening you run just the extraction pump to pull out the pre-heated, pre-pressurized water. You're still at a 3-5x energy return, but the time-shifting has made the cost multiplier more favorable.
My understanding is it's still in research phase, but Fervo is piloting this technique. Like another thread said, they're pre-IPO now, so they've been flooding the renewables media with all these stories. They filed an S-1 recently, but always read the eventual S-3 before considering your investment options blahblahblah.
> Several companies are now building upon existing techniques for accessing geothermal resources by integrating enhanced geothermal systems (EGS) into operations. While conventional geothermal systems produce energy using hot water or steam, pumped from naturally occurring hydrothermal reservoirs trapped in rock formations underground, EGS use innovative drilling technologies, such as those used in fracking operations, to drill horizontally and create hydrothermal reservoirs where they don’t currently exist.
Drilling horizontally doesn’t magically reduce the depth, nor the problem that drilling in to hot rock is like drilling in to plasticine, at least for temperatures worth working with.
In traditional fault hosted (not magmatic) geothermal the convection of the water up the fault brings the thermal energy closer to the surface where drilling depths are economical. This convection heats the surrounding rock and over hundred thousand - million of years brings the background temperature around a large volume at depth surrounding these systems considerably above traditional background geothermal gradients. By drilling into a much larger volume of impermeable hot rock surrounding a very small permeable fault hosted section you can considerably enhance the power potential of a traditional fault hosted geothermal system (the E in EGS). That is what Fervo is doing and why their projects are situated right next to traditional geothermal power plants.
The assumption is that if you can increase drilling efficiencies enough then you don't even need a fault hosted or similar system to bring that energy close to the surface, you can just drill down deep enough to get at similar temperatures. That is a big assumption in the economics.
EGS has been around for at least 15 years. See AltaRock Energy as an example (I’m sure there are others). They started almost 20 years ago.
https://en.wikipedia.org/wiki/AltaRock_Energy
No. Current geothermal projects need very specific geology to work, its very rare which is why geothermal is such a small blip in the overall energy picture. Enhanced Geothermal Systems (EGS), the technique Fervo is using, can create the conditions to be able to generate electricity. The hope is this will greatly expand the number of projects that can be developed.
My understanding is that it's due to better drilling techniques. The industry learned a fair bit from fracking and they're learning more from experience as they apply it to geothermal.
No particular breakthrough, but there's a learning curve and they learn more as they do more. Other industries sometimes work that way, too.
It’s a stupid ploy by oil drillers to try and expand their existing capex to something more future proof. Their theory is to randomly frack the ground and hope superheated water shows up there. It’s probably just going to end up ruining existing groundwater sources.
55% Palantir revenue comes from government contracts and 50% from the US govenment.
With this "are we the bad guys" perspective, I wonder how much of the "evil" they are apparently doing is a result of the current view a majority of people globally have with the current administration?
Though we may find it difficult to separate the two, because it seems leadership and the founders of Palantir are supportive of, and in some ways responsible for, Trump getting elected, but with different leadership using the tools in different ways, would we still consider Palantir the bad guys?
personally yes, i've considered them the bad guys from day one. they have always publicly portrayed themselves as enabling mass surveillance so i'm not even sure why this sudden crisis of conscience, unless the trump administration has finally made it clear to even the thickest-headed of them that mass surveillance is not a good thing.
The intro is a good mentality to how all founders should approach challenge.
I'm surprised by the 2nd bit of the introduction, which initial sounds like the Canadian version of MAGA, strangely protectionist, with secondary reliance on new partners.
The comments pointing out Brock and 1812 is a reminder that US and Canada were once enemies, which I think was an unnecessary approach.
Carney doesn't put out messaging without considering every word.
It's also strange to see a head of state building an audience on youtube. Is Trump setting the playbook for politicians on social media? Perhaps not being on social media much myself, I just don't see it.
I'm Canadian, have lived in the US, and currently live in Australia.
My co-founder and I were both software engineers who got into hardware almost 6 years ago.
We had both played with a bit of arduino or hobbyist stuff in the past, but dove right into EEG and neurostimulation.
Hardware is hard like software used to be hard. When we think of hardware, we're usually caught up just in the electrical chip, transistor, firmware, IO mentality.
Maybe it's just because of what we're building (https://affectablesleep.comm) but we also have to factor in mechanical design, electrical/mechanical interfaces, material science (electrode sensors), etc.
I've learned power management isn't always about, how long does my battery last, but also, how do micro-fluctuations in voltage during use impact other sensors or other operating systems.
Hardware, for us at least, has been a significantly larger footprint than we initially would have expected. It isn't just hardware, firmware, control app. For us it's been the above, plus data-quality tooling, firmware CI, in-factory provisioning and security, the list seems to go on and on.
It's a new challenge, and has been a very fun one. It does remind me a bit of the early days of the web when we'd write our own web-server, DB connection logic, etc etc.
I never would have imagined that OpenAI is sending the full audio of a request to their servers. I had always assumed the audio was transcribed locally and then sent to the server.
The only reason I can think they'd want the full audio is for later model training, which, ok, fair-enough, but this can still likely be done without the limitations of WebRTC.
reply