He's the CEO, and there's now been a few "oh geez some rogue employee made grok say white supremacist stuff, we totally didn't mean for it to say that!" moments.
If the management isn't fixing the problems that led to those events, the management is responsible.
I don't like speech to text in this context because of how often negation words are either missed or inserted, and how commonly "I want <item>, no <ingredient>" and "I want <item> with <ingredient 1>, <ingredient 2>, <ingredient 3>" show up in the fast food ordering context.
That's also the kind of small detail likely to be missed by the human who is only half-listening to the conversation being had with the customer.
At the drive thru, the order was displayed on a large screen as you added items. You could plainly see if it got something wrong, and you could verbally have it make some corrections. If it got totally f-ed up you could just ask for a human.
Having a screen displaying your order is something many drive-through restaurants implemented long ago. It's a useful error detection mechanism regardless of what is interpreting the customer's speech.
We can also say, having observed those implementations for the last few decades, that the system is not always working for one reason or another. In my own experience, when the capability appears to be there, I see my order details about half the time. Some of that is probably humans not pushing buttons they're supposed to, but likely also includes a bunch of technical failure conditions.
When I'm already talking to another human and I don't have a way to visually inspect the order as it's being built, I can ask for confirmation of items and modifiers. I can focus this on areas where there might be more confusion - "did you catch the extra mayo for that?" and so forth.
If the order display system isn't working and an AI is doing the speech interpretation, I have less confidence on where errors might be made or the types of errors that are likely to be made. I wouldn't be able to confidently move forward without getting the AI to read back my entire order (taking a lot of time) or transferring to a human (also taking a lot of time, and now I'm burning the human's time).
From the customer perspective, the happy path is not improved by AI order takers. In the best case, you have basically the same experience as you would with a human order taker. The failure paths are made worse. Responsibility for verifying the AI's accuracy is placed on the customer, and the customer also has to be the one saying "I need a human to intervene." As many errors will eventually require human intervention, the time taken to resolve an error will tend to be longer than without AI as the customer has to get past whatever guardrails are in place to prevent immediate transfer out of the AI flow. The error rate is likely to be higher with AI order takers in general, meaning customers encounter these failure paths more often.
IMO that's a common theme with a lot of AI customer service 'solutions' out there today...from the customer perspective, happy paths are minimally (or not at all) improved, and failure paths shift cognitive load and responsibility from the business to the customer.
Do you think anyone makes the same error when they see a "self cleaning" oven?
There's plenty wrong about the FSD terminology and SAE levels would absolutely be clearer, but I doubt more than a tiny fraction of people are confused as to the target of 'self' in the phrase 'full self driving'.
> Do you think anyone makes the same error when they see a "self cleaning" oven?
How many juries and courts have ruled adversely against self-cleaning oven makers?
Tesla has absolutely lied about its software's capabilities. From the lawsuit that went to trial:
“In 2016, the company posted a video of what appears to be a car equipped with Autopilot driving on its own.
‘The person in the driver's seat is only there for legal reasons,’ reads a caption that flashes at the beginning of the video. ‘He is not doing anything. The car is driving itself.’ (Six years later, a senior Tesla engineer conceded as part of a separate lawsuit that the video was staged and did not represent the true capabilities of the car.) [1]”
To be 100% clear: FSD and Autopilot are both terrible product names that imply promises greater than the products can deliver, and Musk / Tesla have made that worse with statements like those you reference. People have died as a result.
I just disagree that any significant number of people anywhere have thought the 'self' in 'full self driving' refers to the driver.
Musk’s assistant peeked back the muttered and said he had another meeting. “Do you have any final thoughts?” she asked.
“Yes, I want to say one thing.” the data scientist said. He took a deep breath and turned to Musk.
“I’m resigning today. I was feeling excited about the takeover, but I was really disappointed by your Paul Pelosi tweet. It’s really such obvious partisan misinformation and it makes me worry about you and what kind of friends you’re getting information from. It’s only really like the tenth percentile of the adult population who’d be gullible enough to fall for this.”
The color drained from Musk’s already pale face. He leaned forward in his chair. No one spoke to him like this. And no one, least of all someone who worked for him, would dare to question his intellect or his tweets. His darting eyes focused for a second directly on the data scientist.
“Fuck you!” Musk growled.
The "real-time" version looks awful with constantly shifting colors, inconsistently sized objects, and changing interpretations of the underlying data, resulting in what I would consider an unplayable game vs the original ASCII rendering.
The "better" version renders at a whopping 4 seconds per frame (not frames per second) and still doesn't consistently represent the underlying data, with shifting interpretations of what each color / region represents.
Yeah, as interesting as the concept is, the lack of frame to frame consistency is a real problem. It also seems like the computing requirements would be immense—the article mentions burning through $10 in seconds.
You can do this at home on your own computer with a 40x0 consumer GPU at 1-2 fps. You have to choose a suitable diffusion model, there are models that provide sub-second generation of 1024x1024 images. The computing requirements and electricity costs are the same as when running a modern game.
I like the idea behind https://oasis-ai.org/ where you can actually try to take advantage of the 'dream logic' inconsistency of each frame being procedurally generated based on the last one. For example, instead of building a house, build the corner of a house, look at that, then look back up and check if it hallucinated the rest of your ephemeral house for you. Of course that uses AI as the entire gameplay loop and not just a graphics filter. It's also... not great, but an interesting concept that I could see producing a fun dream logic game in the future.
OP here. Thanks for the feedback. I agree that frame to frame consistency is quite bad currently. I did address that in the post, hinting at some of the techniques others have mentioned here, like in/out-painting and masking previous frames.
For me, the exciting parts of this experiment was finding the opportunities and limits of realtime generation, and exploring ways of grounding generated content in a solid yet player controlled world layer.
Because our ability to simulate/render a realistic world in real time using direct equations is still very limited. We’re accustomed to these limitations and often feel “graphics are good enough”. But, we’ll always be decades behind “ILM in real time”.
The AI route has a good chance of moving us from decades behind ILM to merely “years behind ILM”.
>The AI route has a good chance of moving us from decades behind ILM to merely “years behind ILM”.
Firstly: we have very accurate models. But not at real time speeds. Games only have some 30,16, or even 11 ms to render a frame. The techniques we have are faking the real physical interactions that render farms can take minutes or hours to pump out per frame.
Secondly: Not at this performance rate. taking 100ms to render a frame is unaccatable in the concept of a game. Games are already so pressed for time budget; unless some hyper JIT happens we can't take all that budget querying an LLM.
Not OP, but I have long thought of this type of approach (underlying "hard coded" object tracking + fuzzy AI rendering) to be the next step, so I'll respond.
The problem with using equations is that they seem to have plateaued. Hardware requirements for games today keep growing, and yet every character still has that awful "plastic skin", among all the other issues, and for a lot of people (me included) this creates heavy uncanny-valley effects that makes modern games unplayable.
On the other hand, images created by image models today look fully realistic. If we assume (and I fully agree that this is a strong and optimistic assumption) that it will soon be possible to run such models in real time, and that techniques for object permanence will improve (as they keep improving at an incredible phase right now), then this might finally bring us to the next level of realism.
Even if realism is not what you're aiming for, I think it's easy to imagine how this might change the game.
You're comparing apples to oranges, holding up today's practical real-time rendering techniques against a hypothetical future neural method that runs many orders of magnitude faster than anything available today, and solves the issues of temporal stability, directability and overall robustness. If we grant "equation based" methods the same liberty then we should be looking ahead to real-time pathtracing research, which is much closer to anything practical than these pure ML experiments.
That's not to say ML doesn't have a place in the pipeline - pathtracers can pair very well with ML-driven heuristics for things like denoising, but in that case the underlying signal is still grounded in physics and the ML part is just papering over the gaps.
The question was "why does it feel more real", and I answered that - because the best AI generated images today feel more real than the best 3D renders, even when they take all the compute in the world to finish. So I can imagine that trend going forward into real-time rendering as well.
I did not claim that AI-based rendering will overcome traditional methods, and have even explicitly said that this is a heavy assumption, but explained why I see it as exciting.
I think we'll have to agree to disagree about well done 3D renders not feeling real. Movie studios still regularly underplay how much CGI they use for marketing purposes, and get away with it, because the CGI looks so utterly real that nobody even notices it until much later when the VFX vendors are allowed to give a peek behind the curtain.
e.g. Top Gun Mavericks much lauded "practical" jet shots, which were filmed on real planes, but then the pilots were isolated and composited into 100% CGI planes with the backdrops also being CGI in many cases, and huge swathes of viewers and press bought the marketing line that what they saw was all practical.
I find it odd that you're that bothered by uncanny valley effects from game rendering but apparently not by the same in image model outputs. They get little things wrong all the time and it puts me off the image almost instantly.
>Hardware requirements for games today keep growing, and yet every character still has that awful "plastic skin", among all the other issues
That's because the number of pixels to render onto keep growing. Instead of focusing on physically based animations and reactions, we chose to leap from 480p to 720p overnight, and then to 1080p in a few more years. Now we quadrupled that and want things with more fidelity with 4x the resolution of last generation.
> images created by image models today look fully realistic.
Because they aren't made in real time (I'll give the BOTD for now and say theya are "fully realistic". Even this sample here claims 100ms. Rendering at 6-7 seconds per frame isn't going to work for any consumer product at any point in gaming history.
>Even if realism is not what you're aiming for, I think it's easy to imagine how this might change the game.
not in real time rendering. I am interested to see if this can help with offline stuff, but we're already so strapped for performance withoout needing to query an "oracle" between frames.
as of now, I'm not even that convinced by the Nvidia 5X series of frame interpolation (which would only be doable by hardware manufactureres).
What relevance does their plausible earnings via the offense have to the fine for the offense?
The harm suffered by the people whose privacy was violated is still there regardless of how much money was made through the violation.