[ad_1]
Separating fact from fiction. The past few weeks have simply gone on to emphasise the point that it is all fun and games to envision a world layered with artificial intelligence, until it isn’t. Re-emphasise, rather. How do you really figure out the difference between a real photo and an AI generated image (in context of it being shared, a fake image) when you scroll through your social media feeds. Nigh impossible. More so, since text-to-image tools have evolved rapidly in terms of realism. Generated videos too, now, to add complexity.
There are also many, many of them. Easily available, on phones and the web. Often free (and some have paid subscriptions, unlocking even more tools for realism). Taylor Swift. Drake. Before that, Rashmika Mandanna, Donald Trump, Elon Musk Ryan Gosling. This list will only get longer. Illustrates a level of realism artificially generated (AI) images are now capable of. Nevertheless, this week we may have taken a few big steps forward, for countermeasures.
Meta. OpenAI. Google. Adobe. The tech giants we need to keep an eye on in the coming months. I’ll summarise.
Meta says they’re ready to include labels on posts on Facebook, Instagram and Threads, with AI generated content. Basically, what they’re saying is, systems will be in place to look closely for otherwise invisible signals in images being shared. They already do include watermarks and have visible markers on media generated using the Meta AI tool. Basically, the IPTC Photo Metadata Standard by the International Press Telecommunications Council and invisible watermarks, in line with Partnership on AI (PAI) best practices.
But for that to work, the source point for AI generations, will have to put measures in place. Industry standards for visible and invisible markers with each generative AI creation, will need to be placed by Google, OpenAI, Microsoft, Adobe and Midjourney, or any generative AI tool that’s out there. That would allow Meta (and hopefully more platforms join in) to use these signals for labels accompanying posts featuring AI content.
OpenAI says they’ll begin to add watermarks to content created using their image generator, Dell-E 3, whether directly in the tool or using the popular ChatGPT chatbot. This will support standards from the Coalition for Content Provenance and Authenticity (C2PA).
Google is also joining the C2PA to support the push for “content credentials” with generated content. Tech and photography companies including Adobe, Microsoft, Intel, Leica, Nikon, Sony and Amazon Web Services are part of the C2PA. This is essentially an open technical standard that allows publishers, companies, and others to embed metadata in media for verifying its origin and related information.
As it often is, some tech companies foresee what is round the corner, better than others. Or even if they don’t, at least there’s a willingness to move forward. It was in July last year, the tech giant had talked about transparency for digital content. The timing was important, as Adobe’s Firefly generative AI platform found significant traction, standalone and integrated within its popular creative apps including Photoshop and Lightroom. They’d then called this information overlay ‘nutrition labels’. Now they say, these content credentials are also part of any generations using Adobe’s apps on the Apple Vision Pro platform.
I’m most surprised (in a good way) by camera makers and their willingness to play ball in this battle to separate real photos from generated images. In October last year, Leica introduced the world’s first camera, the Leica M11-P, with ‘content credentials’ integrated. That meant adding a layer of authenticity at the point of capture itself. Sony will also add the metadata functionality into its Alpha 9 III, Alpha 1 and Alpha 7S III cameras, via firmware updates. Nikon too, for its next line of cameras. All this considered in entirety, we may just be getting somewhere. May. Just.
Comprehend: This random thought crossed my mind. Twitter, now X, began selling its “Blue” subscriptions with the promise of boosted tweets. Replies specifically. That later became Smallest, Larger or Largest depending on Premium subscription tier and how much money you wanted to send Elon Musk’s way as he tries to achieve his dreams of keeping X afloat. Your posts boosted. My posts boosted. Delightful. But think about it — once the Premium subscription base reach a certain number (what must that number be?), wouldn’t the boosting for replies by each paying subscriber, get cancelled out by another paying subscriber? And so on? For a post on X, how many replies beneath it can be artificially given priority?
INTELLIGENCE
It was getting beyond perplexing. Even though Google had Bard available for many months now, it had no entry points on an Android phone at all. In the same time, Microsoft made it easier for users to access Copilot, via the Edge browser, the Copilot app and the Bing app. OpenAI had their own app. Perplexity even added a neat widget into theirs, for home screen visibility. Google expected a user to open Chrome, type in bard.google.com and then begin with the burning questions. It was never going to pan out as intended. Things are now moving, and in the style that Google seems to adopt for AI announcements – dignified, yet grand (remember the Gemini models, late last year?)
It is now an app on Android phones, can be the new assistant on Android phones, and part of the Google app on iPhones. But there’s more to it, than just apps. Underlying this chatbot on phones and computers are two of the three Gemini models Google had unveiled in December (Refresh your memory). Those are Gemini Pro and Gemini Ultra. And that’s the peg for an optional subscription model too. If you do choose to, the Google One AI Premium plan will demand you part ways with ₹1,950 per month – that’ll among other things such as 2TB cloud storage, unlocks access to Gemini Ultra model for your conversations with AI. If you don’t want to subscribe, there’s little you’ll be missing out on (Gemini Pro isn’t a step down in terms of experience).
Pro tip: Want to spend? Google’s pricing is at par with Microsoft Copilot Pro ( ₹2,000 or around $20 per month) and OpenAI’s ChatGPT Plus ( ₹1,999 or around $20 per month). Are you ready to pay for an AI chatbot? If yes, you may ideally find more value picking the Google One AI Premium plan, because it also includes 2TB cloud storage and more. For Microsoft and OpenAI’s subscriptions, you pay for just a chatbot.
I’ve often heard about “personalities”, and I don’t pay much heed to it. That’s all claptrap. Ignore. Generative AI’s job, is simple – give accurate and detailed responses, based on what is asked of it. Be it text search, voice search or a text prompt to generate an image or video or music clip. It is that simple. Choose one, which you feel works best for your queries and tasks. I don’t want it to be creative, or try to use its own ingenuity. Straight up, accurate facts.
On that front, I decided to have the same conversation with Gemini (Pro, not the paid Ultra) and Microsoft Copilot (not the paid Pro tier), and it turns out that as it averages out, there’s very little difference between the responses. Gemini gave me more detail (See more, on myX post; a little bonus, Gemini’s reply is interesting) when I shared a photo of the Gemini constellation with it than Copilot (no coincidences here, I’d guess), but when I asked them which coffee they each prefer, they were equally factual saying they’ve no specific preference, but went on to detail every bit possible from search about coffee and variety. Now that’s more like it.
COLLECTIVE
The moment I heard this, the first reaction was – this can only be good news for all users, be it for the consumer app, or business users. Proton, a company whose services I recommend more than others purely because of the strong data privacy foundations, has widened the scope of its Proton Pass password management platform. It’s now rolling out for businesses too. Just weeks after Proton did something quite rare, which is reduce the subscription cost from $3.99 per month to $1.99 per month (we only seem to partake in subscriptions becoming costlier, but never the other way). Proton said a combination of input and running costs have reduced. Honest.
Read: Proton Pass for business will benefit from lower prices and Sentinel protection
Proton has smartly retained feature parity between Proton Pass and Proton Pass for Business. An able to generate random combination passwords to secure online accounts, integration across web browsers and on smartphone platforms including Android and iOS to enable sign-in autofill, two-factor authentication code generation, built-in encrypted notes and the Proton Sentinel program that uses AI and human intervention to respond to suspicious indications.
The following article was originally published in this week’s Wired Wisdom. Click here to subscribe.
[ad_2]
Source link
