Listen to the discussion from this article here on The Ad Stack On the Air:
Listen to “The Ad Stack On the Air” - a discussion and analysis of this edition by personable AI DJ’s. From Notebook ML. Also available as a podcast on Spotify, and others players.
My friend Tim O’Neill from Australia’s own Time Under Tension threw this story into my feed this week.
Source: Reuters, August 30, 2024
Ask Claude: Amazon turns to Anthropic's AI for Alexa revamp
“Remarkable Alexa / Claude / Alexa?”
1. Alexa becomes another $10 sub.
In a plot twist that surprises absolutely no one except my family, Amazon has decided that what their Alexa voice assistant really needs is a brain transplant. The online retail giant, not content with merely accidentally listening to your every word, is now set to make Alexa smarter - or at least putting on a better show.
Enter "Remarkable" Alexa, the soon-to-be-released upgrade that promises to turn your home into a sitcom of AI-powered hijinks. And who's the star of this digital dramedy? None other than Anthropic's Claude, the AI chatbot that apparently impressed Amazon so much, they decided to outsource Alexa's intelligence altogether.
Why Claude, you ask?
Well, it seems Amazon's homegrown AI’s lagging response times put us to mind of the comfort provided by a 28.8 baud modem - taking a leisurely 6-7 seconds to respond. In the fast-paced world of asking your speaker what the weather's like, that's practically an eternity.
Does Amazon truly give a sh-t about your Alexa experience?
Maybe. Here are two items they probably find more prescient:
1. There’s only so much black plastic hardware you can sell per household.
I have a large screen in the kitchen, two small screens that make ideal alarm clocks, and one of the tall cylinders for Bluetooth audio with my desktop PC. Three are piled in the basement … somewhere. A dead end revenue stream.
2. Performance is lacking.
Siri on my watch and phone responds on the first request. Alexa requires 2-3 attempts, and she really doesn’t want to play Spotify music. She much prefers I subscribe to Amazon Music. UX is stale and boring.
The Fix
A Claude refresh establishes a new software revenue stream. (Yet another) $10 a month subscription instead of free.
Amazon invested $4 Billion in Claude last year. Nothing wrong with sweetening that investment a bit.
The upside - Gone are the days when "Play white noise" turned into "Display my browser history to all of my neighbors."
The downside - aside of the additional monthly cost, Claude in it’s present configuration does not offer the freshest data (nothing usually within the past year). This is due to keeping A.I. in a closed system so it’s not running amok launching ICBM’s via connecting the data hose to the live net. It will be interesting to see their approach here.
Remarkable Alexa Better Be Remarkable
What wonders await in this brave new world of "Remarkable" Alexa? Prepare to be amazed as your AI assistant:
Carries on conversations that make you question your social life
Offers shopping advice, because who needs human friends when you have a corporation's AI telling you what to buy?
Aggregates news stories, ensuring you never miss out on the latest celebrity gossip or '“DEAD OR ALIVE” game fodder.
Automates your home to the point where you're essentially just the organic component in a smart home experiment.
Amazon, in its infinite wisdom, hopes this upgrade will finally make Alexa profitable. Because nothing says "sound business strategy" like charging for a service people barely used to set multiple cooking timers.
From Reuters:
Amazon hopes the new Alexa will also be a supercharged home automation hub, remembering customer preferences so that, say, morning alarms are set, or the television knows to record favorite shows even when a user forgets to, they said.
The company's plans for Alexa, however, could be delayed or altered if the technology fails to meet certain internal benchmarks, the people said, without giving further details.
As we hurtle towards the 2024 holiday season, the tech world watches with bated breath. Will "Remarkable" Alexa revolutionize our homes, or will it just be another gadget that judges our music taste and reminds us we're out of milk?
2. Adobe's Firefly Makes It to the Top of the Generative Video Shitshow.
In a world where content is king and attention spans are shorter than a TikTok dance, Adobe has decided to play god with its latest creation: the Firefly Video Model. Launched in March 2023, Firefly started as a humble image generator, but like a steroid-infused butterfly, it's metamorphosed into a video-editing behemoth that threatens to make human creativity as obsolete as thumb drives.
With over 12 billion images and vectors already conjured up by its community, Adobe isn't content with just still images. No, they're diving headfirst into the choppy waters of video generation, promising to turn all of the mildly-talented into the next Spielberg – or at least the next viral cat with no children video creator.
“Just shoot it in post.”
Set to grace us with its presence in beta later this year, Firefly Video is being touted as the Swiss Army knife of video editing. Need B-roll of a sunset over a mountain? Just type it. Want to extend that clip of your dog looking cute for a few more seconds? Firefly's got you covered. It's like having a genie in a bottle, except this genie runs on gobs of electricity and doesn't judge you for your sh*tty ideas.
Adobe claims they've worked closely with the video editing community to develop this AI wunderkind. The result is a tool that promises to fill gaps in timelines, remove unwanted objects (goodbye to the unattractive), and smooth out workflow. Just shoot it in post.
Prompt: Slow-motion fiery volcanic landscape, with lava spewing out of craters. the camera flies through the lava and lava splatters onto the lens. The lighting is cinematic and moody. The color grade is cinematic, dramatic, and high-contrast.
Throwing Your Inner Artist a Solid
But wait, there's more. Firefly isn't just about fixing your mistakes. It's here to unleash your inner artist – or at least the artist you think you are. With text-to-video generation, you can create entire scenes from a prompt. “360-degree pan of squirrels reenacting the French Revolution," you muse? Voilà! It appears before you can exclaim "Alvin and Les Misérables.
"If you’re looking for a key differentiator (as am I) one actually exists.
For those worried about the ethical implications of AI-generated content, fear not. Adobe assures us that Firefly is trained only on content they have permission to use.
Do the possibilities seem endless? Who knows. They all seem the same more-or-less; just in differing states of disrepair. If this LLM / LMM is trained on legit content, will the quality of results take a hit? Who knows. This is key. Presently stock houses including Getty and**Adobe will license “cleared” gen-ai stock imagery you can use on client work. Adobe’s Firefly upgrade goes beta this Fall so we should soon know.
**Getty stated here today this is not true.
Adobe's new toy isn't just about creating content; it's about saving time. In a world where clients want everything yesterday (charge them for it), Firefly promises to be the Red Bull guzzling assistant editor ideally suited for agency deadlines.
Original reference footage perhaps supplied by client.
Created footage generated based on the original supplied clip (above).
Prompt: Detailed extremely macro closeup view of a white dandelion viewed through a large red magnifying glass
With the Firefly Video Model you can leverage rich camera controls, like angle, motion and zoom to create the perfect perspective on your generated video.
Final edited sequence. Crazy.
As we stand on the precipice of this brave new world of AI-assisted creativity, one can't help but wonder: Is this the dawn of a new era in video production, or the beginning of the end for human imagination? Will we soon see Oscar acceptance speeches thanking themselves and their software instead of family members?.
Some Sage Advice (revised 9/16/24 - it was late)
A friend once told me to “pick two” apps as I was trying to learn everything in sight. Great advice. I chose Photoshop and Premiere. Among the many Generative AI choices your selection(s) with which to invest your time really depends on your goals, needs, and perhaps some corporate requirements.
Thank you for the read.
Feel free to share comments below. If you enjoyed this read, please whack the bejabbers out of the heart icon …
…at the top or bottom of the page so others can find this article. Re-stacking is great too.
Have a great week people!
BTW - agreed on the Alexa comment (no one will pay). Mine barely work. I might go for paying for a regular Claude sub and get the Alexa upgrade at no extra cost. Claude is pretty awesome on its own. Plus it's free at the level I'm using. Yeah, never mind.
You KNOW I have to comment on this, and with many words in all caps for emphasis.
First - no one is going pay for Alexa. NO ONE.
Second - Adobe's new video tool is trained on their own library and public domain content (1:16 in their release video). That's a wild world of stuff that the copyright MAY have been given away (you never really know), but that doesn't answer what it's of. I can take a picture of Brad Pitt and upload it to Creative Commons giving away my right to the copyright, but Mr ex-Jolie certainly did not give up his right to publicity.
Third - Of the major stock libraries ONLY Adobe accepts AI generated images (at this point it's 1/4 of their image library), which means they're accepting MidJourney/Stability/DALL-E and all of the other scraped sources, AND also using that to train their tool - on top of the unknown "public domain" content. https://martech.org/legal-risks-loom-for-firefly-users-after-adobes-ai-image-tool-training-exposed/. Shutterstock doesn't accept third party GenAI content, but they do recycle what you've created with their tool back into the library (so your prompts and creations are up for anyone to license with no attribution to you). Getty doesn't accept AI content nor make anything created in their generator public.