What are your favorite features? I recently downloaded it and also use Codex CLI and GitHub Copilot in VS Code but I don't really know what specific features it has others might not have.
The UI is better - they box the specific types of actions the orchestrator agent takes with a clear categorization. The standard quality of life shortcuts like type a number to respond to an MCQ are present here as well. They use specialized sub agents such as one with big context window to find context in the codebase. The quotas appear to be much more generous vs CC. The agent memory management between compacting cycles seems to have a few tricks CC is missing. Also, with 3.0 Flash, it feels faster with the same level of agency and intelligence. It has a feature to focus into an interactive shell where bash commands are being executed by the orchestrator agent. Doesn't feel like Google is trying to push you to buy more credits or is relying on this product for its financial survival - I suspect CC has some dark patterns around this where the agents runs cycles of token in circles with minimal progress on bugs before you have to top up your wallet. Early days still.
It’s unclear to me wether that would give some access to a token quota or if it would just be like any other « Sign in with … ». In all cases I am currently developing an app that would greatly benefit from letting my users connect to their ChatGPT account and use some token quota.
The issue I see is that for certain apps, such as one I am currently working on and hope to publish soon on iOS, is that they do require a lot of maintenance once published even if there were no server costs. Given the amount of work I already put in it and how much more will be necessary even just to keep the app correctly running in the future, I don't really see what other monetization approach would make sense for me. Actually, I would even argue that selling an app without a subscription might (sometimes) be setting wrong or blurry expectations: if a user accepts to pay today a single time, how long are they expecting updates for? Will it only be basic bug fixes or also major new features? With a subscription, I feel like at least if they are unhappy with my app, they won't really have lost anything and can just unsubscribe, since they had basically accepted, IMO, that the money they put in my app each period of time is only for the service and potential updates in that small period of time and not future changes.
I will say I love this feature on my iPad (touch the screen and hold without moving, the video will be 2x until you let go) but I wish it could be turned on on a per-device basis.
It is streamed. They also offer this as an API for developers [0].
I just wish they'd offer more languages as it currently is English only. It can speak German and Spanish, but with a very clear English accent, which makes it kind of funny and makes one think about why the accent sounds so real.
Just tried now, and it was pretty good. Some slight accent that I wasn't able to detect (Minas Gerais?). Similar / better than the artificial voice of other assistants in pt-br. I'm using Sky's voice, FWIW.
When I tried it right when it was released, it was some really heavy accent from Piracicaba. People there talk like that but ChatGPT seemed to be doing a forced impression of that.
It seems similar to what other commenters are saying about it taking in German with an American accent.
It's no big deal but Siri/Google/Alexa voices are better normalized, IMHO. I don't mind heavy accents at all in humans but they get annoying after a while for a voice assistants, especially if it's not my accent :)
I tried it now but it doesn't seem to be working for me. Maybe related to the outage yesterday (is it still going on?). It just listens but never replied.
...does this directly relate to the high operating costs of LLMs-as-a-service, if for every request they have to run n-many redundant LLM requests? So if they could improve things so that a single prompt/request+response has a higher chance of being high-quality they wouldn't need to run alternatives?
It can make it more expensive if that option becomes popular.
But I think in most cases batching is actually the biggest _improvement_ in terms of cost effectiveness for operators, since it enables them to use the parallel throughout of the graphics device more fully by handling multiple inference requests (often from different customers) at once. (Unless they work like Bard by default).
I thought the char-by-char teleprinter thing was just an effect (y’know, for user-engagement and to make the interaction feel more genuine) - and that these systems just return output in buffered blocks/pages or whatever-it-is that they wired-up their network to do.
It’s not like DALL-E outputs pixels in scanout order - or in brushstroke order (…er… or does it?)
It's not an effect at all. It calculates and outputs one token at time. The algorithm requires all previous tokens in order to output the next one. DALL-E is a totally different algorithm. It does not have a scanout or brushstrokes.
I wonder the same thing about cellular iPads sometimes because I have a hard time seeing a case where you'd have an iPad with you but not your phone to make some connection sharing. My supposition is that this might just for convenience: no need to worry about your phone battery life, for instance. I'm also curious about whether I'm missing something tho.
Most people barely understand the differences between WiFi and cellular, and the number that can start a hotspot and connect without it being a 'stressful computer problem' is a lot fewer than you would expect. Like, painfully fewer.
If we're talking about the Apple ecosystem, you don't need to start a hotspot.
When you have an iPhone and MacBook nearby that are on the same Apple ID, the Mac shows a little notification "Connect to the Internet through <phone name>?" with the default button being OK, which instructs the phone to turn on the hotspot and the Mac to connect to it.
The phone also just shows up as another Wi-Fi network in the Wi-Fi menu alongside all the other networks, even if hotspot isn't on, using the same mechanism of letting the Mac remotely configure the hotspot.
reply