Wired Wisdom: No escaping AI videos, but are we ready to answer tough questions about reality? | BetaVersa - BetaVersa
March 28, 2024

Wired Wisdom: No escaping AI videos, but are we ready to answer tough questions about reality? | BetaVersa

8 min read

[ad_1]

In our numerous conversations over the past 12 months, we have touched upon generative AI’s steady walk towards realism. Within that we’ve pointed out how text to image tools are now more realistic (largely gone are hands with six fingers; now attention is on skin tones and dynamic range). The one thing that lagged comparatively was text to video, or any sort of AI generated video. While very much possible, results were, well, easily identifiable. That’s set to chance. Rapidly. OpenAI’s Sora.

FILE - The OpenAI logo is displayed on a cell phone with an image on a computer monitor generated by ChatGPT's Dall-E text-to-image model, Friday, Dec. 8, 2023, in Boston. (AP)
FILE – The OpenAI logo is displayed on a cell phone with an image on a computer monitor generated by ChatGPT’s Dall-E text-to-image model, Friday, Dec. 8, 2023, in Boston. (AP)

One minute long videos based on a text prompt, and to be honest, very realistic looking ones (including a human’s facial elements and the ambient scene). It’ll be available to all as part of OpenAI’s products in the coming days. I was reading through the technical report (you must read too, here), and they make it clear the inspiration comes from large language models and generalist capabilities. That’s broader scope that what has been currently tested, including recurrent networks, generative adversarial networks, autoregressive transformers and diffusion models.

Hindustan Times – your fastest source for breaking news! Read now.

Who can access Sora, now? “Sora is becoming available to red teamers to assess critical areas for harms or risks. We are also granting access to a number of visual artists, designers, and filmmakers to gain feedback on how to advance the model to be most helpful for creative professionals,” says OpenAI.

The fact that it can do complex scenes, handle multiple subjects and elements in the same frame, reproduce a sense of motion (be it a tracking shot of a person walking, or what seems like a drone shot surveying an area) and return realism with details, will simply make it more difficult to distinguish AI videos from videos shot in the real world. I was quite intrigued by the prompt “the camera follows behind a white vintage SUV with a black roof rack as it speeds up a steep dirt road surrounded by pine trees on a steep mountain slope, dust kicks up from its tires, the sunlight shines on the SUV as it speeds along the dirt road, casting a warm glow over the scene…”. Simply put, the more detailed you get with the text prompts, the richer detailing models will generate (or at least that’s the promise).

Then there’s the other side of the spectrum. The prompt “Reflections in the window of a train traveling through the Tokyo suburbs” has to be seen to be believed. What you momentarily see in those reflections, as if real passengers are in that train, is well, real.

Reflections on a train (SORA-generated image)
Reflections on a train (SORA-generated image)

But OpenAI is also tempering expectations. The current model has weaknesses. It may struggle with accurately simulating the physics of a complex scene, and may not understand specific instances of cause and effect. For example, a person might take a bite out of a cookie, but afterward, the cookie may not have a bite mark. The model may also confuse spatial details of a prompt, for example, mixing up left and right, and may struggle with precise descriptions of events that take place over time, like following a specific camera trajectory.

Text to video AI space will be the one to watch this year. Alongside Sora are Google’s Lumiere, Runway and Pika. As will be any (it is not a matter of if, but when and how) measures to watermark or identify AI videos from real ones. We’d spoken about it extensively last week –labels and watermarks becoming the go-to choice to identify generated content. Adobe set off on this mission much before everyone else, but with the likes of OpenAI, Meta and Google now more than willing to take part, we may have a more responsive beginning to a genuine battle against fake content passed off as real, on social media.

As my colleague Binayak Dasguptaexplains in this piece about Sora’s capabilities, one of Sora’s stated capabilities — extending videos or creating motion from a still frame — is an ability with dangerous implications, especially if such tools are broken to put people doing things that they did not do.

JOURNEY

The timing was as pristine as it gets. It had been six months since I last spoke with Xiaomi India president Muralikrishnan B (it was a stressful time then, and it showed). Soon after we caught up again (he’s got his smile back, just as Xiaomi’s re-discovered its form), the Xiaomi 14 teasers dropped. This is an important moment in the company’s journey in India, which incidentally is 10-years “young” as Muralikrishnan calls it. The Xiaomi 14 will be a pivotal attempt to take their seat at the table of flagship Android phones. They’d one foot in the door a few years ago, but many factors combined, that door slammed shut.

Read: Push towards premium is a journey, and we’ll persist: Xiaomi’s Muralikrishnan B

There have been many changes as Xiaomi changed focus to get back in the game, something I’ve illustrated in my piece, with accompanying market share numbers to illustrate a certain degree to which they’ve worked. A change in the very approach to their presence in the Indian market, has truly unshackled them. “There used to be a time when market share numbers and volumes used to be super important for us. But now I think it is more important to be the preferred choice of customers and there is a lot of work for us to do,” Muralikrishnan told me. It’s taken the pressure off, energies better spent on curating experience (retail store changes, being one) and product portfolio.

Samsung and OnePlus largely dominate the space above the 40,000 price point, with Google’s Pixel phones and indeed Vivo’s efforts with the camera (they’ve the Zeiss partnership) figuring prominently. We don’t have much idea about the Xiaomi 14 pricing yet, but it’ll be a rather unique take for a flagship proposition. I’m curious about the screen size (that’s 6.36-inch) in a space where 6.7-inch or thereabouts are the most found viewing canvas. Also interesting a choice, because Xiaomi had the 14 Pro as an option too, but chose to hold back for now.

Muralikrishnan B
Muralikrishnan B

I asked Muralikrishnan about the appetite to succeed in the premium spaces, and he points out they already have had significant success with QLED and OLED TVs. That is true. For phones, the realisation is it’ll be a multi-year journey, one which they’ve the stamina and intention to see to its logical next step. Indicative of a company that’s more focused and realises its strengths – they have no intention to launch any PCs or laptops in the Indian market. At least not this year. The intention is to build with tablets as versatile devices, meant for more than one use case. And why wouldn’t they, after the Xiaomi Pad 6 (read our review) has had some success, andHyperOSgives them a software foundation to build the successors with.

Comprehend: Another quick conversation about X. Oh joy. Last week was about how much boost is genuine boost (relevant for those who’ve been cajoled into paying a Premium subscription to Elon Musk). Now, there’s data about an incredibly popular (at least in the US) Super Bowl LVIII (yes, that one with Taylor Swift in attendance, and Tom Brady screaming at his coach). X didn’t hold back, pointing to interaction records (31% more impressions, 41% more posts than last year and so on). Beyond the posturing, and more worrying for Musk (and indeed advertisers) is cybersecurity firm CHEQ’s founder and CEO Guy Tytunovich tellingMashablethat as much as 75.85 % of traffic from X to its advertising clients’ websites during the weekend of the Super Bowl, was fake. In other words, bots. How many, can always be up for debate. Remind me, didn’t Elon buy Twitter to get rid of the bots? In that case, even one, is one too many.

EXCLUSIVE?

X-Box games coming to PS5
X-Box games coming to PS5

It is finally happening. With big gaming studio acquisitions (such as the long-drawn Activision Blizzard one) now done, and yet Xbox Game Pass subscriptions and Xbox console sales slowing down, Microsoft has signed off on a strategy shift that’ll now put gaming titles that were thus far exclusives to Xbox, on rival gaming platforms. For now, four games, will make their way to the Sony PlayStation 5 and the Nintendo Switch. But they’ve not revealed the titles just yet – perhaps, soon? Which titles would you bet on?Sea of Thieves?Forza Motorsport?Hi-Fi Rush? I wouldn’t expect Microsoft to open up the exclusivity padlock completely, just yet. Think of this as a trial, tentative steps forward, whilst diluting the attractiveness of Xbox or even the gaming PC, as little as possible.

Perhaps this announcement for multiplatform had to be officially confirmed because there were murmurs on the internet, for a while now. The ambiguity (whether it’s happening or not) or speculations about specific game titles, may have done more harm than good. Also, Microsoft’s official numbers indicate Xbox Game Pass subscriptions are not growing as fast as they’d have expected. In early 2022, official numbers pegged the subscription base at 25 million. This week, they’ve confirmed that number is at 24 millions. That’s 9 million new Game Pass subscribers in two years.

Phil Spencer who is CEO for Microsoft Gaming, in the Official Xbox podcast (you can see ithere), made three very interesting points. First, the decision that’s been taken to unlock exclusive titles has been made keeping the long-term health of Xbox as a platform, in mind. And second, he believes gaming as a space is placed at an interesting point in time, a time that’s ripe for Microsoft to use what some of the other platforms have right now to help grow the Xbox franchise. Third, and crucially the one that’s making a case for the second, there will be fewer console exclusive game titles in the coming decade. A more wholesome gaming ecosystem awaits? We’ll see.

Remember the Activision and Blizzard game library? Microsoft says Diablo IV will be available for all Game Pass subscribers from March 28. More titles, after that. It is, in theory, a strong foundation for Microsoft to work with and for Xbox Game Pass subscribers on the console or PC, to draw value from. They need it, if new subscriber trajectory has to improve. Activision, Bethesda, Blizzard, King, and Xbox Game Studios. The collective library is massive. But are gamers seeing value?

The following piece was published in this week’s edition of Wired Wisdom. Subscribe here.

[ad_2]

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *

Copyright © All rights reserved. | Newsphere by AF themes.
Verified by MonsterInsights