Amazon New AI Models 'NOVA' Stun The Entire Industry!

[@r9999t]

Our closed source closed weights model is about as good as Llama... Give us your money. Wait, WAIT, Where are you going???

[@isthatso1961]

Apparently these billionaires are all playing a game of catch up with one another; all betting that they'll be the one to be our almighty AI fuedal lords. Time will tell how this turns out.

[@NeoRelic-o8p]

It's an ai arms race. We will continue to accelerate into whatever future awaits us

[@randomdude2582]

huh

[@noahleaman]

IMHO, playing catchup in an arms race that is barely two years old is not much to catch up on - not as much as CEOs of newer companies would like their investors and wide-eyed fan bases to believe.

[@thechosenwon6762]

Yes Sir I know who my betters are now I'll get back to plowing the field...oh wait the robots are doing it now

[@DailyTuna]

It’s OK because open source will be the rebels and we will fight the ugly nerd billionaire overlords.😂

[@QuantumXdeveloper]

GPT remains the leader, with every large language model benchmark nearing saturation unless there is a significant increase in benchmarking compared to human. As we are aware, benchmark tweaking is common, so there are no surprises there.

[@CamAlert2]

Amazon is 6 months late to the party.

[@noahleaman]

Check back in five years to see who was too late.

[@brianmi40]

Doesn't really matter. competition is good, more people working the "problem" is awesome. it's simply a new contribution to the exponential curve that is PROGRESS and taking us to AGI and beyond.

[@brianmi40]

Also, IBM doesn't have even a single multi-modal model, despite WatsonX released a year and a half ago. Late doesn't mean necessarily last place...

[@farmanguliyev]

still pretty new industry

[@ric3354]

6 months?? No no...2 years

[@sokgans]

Amazon will always be ' The Chum Bucket' of the tech industry.

[@jonanddy]

That’s the best reference I’ve ever heard 😂😂

[@djstraylight]

Guess Amazon has been cooking. They got some compute, so not surprising. I love they benchmarked their image model on the least 'nice' image models but competition is awesome.

[@Dis3spectful]

last time it shocked the industry, this time it stunned it. The industry has been through a lot lately

[@saeedneuro]

😂

[@seylaw]

What about their partnerships with Anthropic? On the one hand, AWS is an investor there, on the other hand, they are now trying to compete directly with them.

[@MaJetiGizzle]

The Microsoft plan, but Amazon/AWS’s version of it.

[@NathanIslesOfficial]

Why not when Ai is everything now

[@Haveuseenmyjetpack]

Healthy competition between two groups of individuals, one working for Anthropic and another with Amazon. Now they all have the same tools.

[@mrchongnoi]

My question also!!

[@jonanddy]

When you want innovation you want competition

[@crazytrain86]

"Stuns" tracker increments again lol

[@timm7128]

What a boring presentation by Amazon lol. Hey here are these models they are generally on par or worse than all the stuff you already use lol

[@maximecaron3133]

But much « cheaper »

[@QuantumXdeveloper]

​@@maximecaron3133 and >

[@noahleaman]

AWS itself is “boring” if you are expecting to see product demos. What’s not boring to me though is that AWS is the largest tech platform there is, and companies big and small are already onboarded with several AWS offerings to make their products/services function. For all those existing customers, being just a few clicks away from providing competitive access to useful AI models on the world’s leading internet backend technology provider’s platform is not boring to them. Dog and pony shows are fun though.

[@brianmi40]

Doesn't really matter. competition is good, more people working the "problem" is awesome. it's simply a new contribution to the exponential curve that is PROGRESS and taking us to AGI and beyond.

[@person-jw7vb]

Have you read the new verge article about open ai's shipmas ?

[@NostraDavid2]

12 days of OpenAI!

[@person-jw7vb]

@NostraDavid2 yes !

[@alertbri]

Nice to see them in the zone of all models... How will they differentiate themselves?

[@scottphardin]

Extensive relationships with large corporations who are already heavily relying on AWS

[@noahleaman]

They are AWS… that’s the difference.

[@TDVL]

Cheaper, faster, embedded in AWS. Question is how will the others compete in the Enterprise space.

[@scottphardin]

@ yup. IBM is the only other player even kitted out to compete in the enterprise space and they’re miles behind Amazon.

[@TheKevphil]

The industry is reeling, some have even fallen and can't get up! Oh, the humanity!

[@T1tusCr0w]

I guess Nova pro will be on Alexa soon?

[@slawinsky8951]

That's something we've all been waiting for. Weren't they supposed to announce Alex's new AI capabilities last October?

[@brianmi40]

@@slawinsky8951 Not looking forward to a surcharge to get it. I have other options already so it doesn't add much real value worth paying for monthly (have Ray Ban Meta's, Rabbit R1, Google Home mini and like most do 96% of all prompts from my PC, in my case, using Perplexity as the Chrome toolbar add-on, which is simply fantastic and ready to comment on any Web page or domain you're on...)

[@tilmanscha4285]

mr aigrid. I ask myself this one question which of the models/companies provide the ensured possibility to own and use it in a commercial way what the models combined with the person who uses it , create, I think such an offer will make such a model the most favorite for the people

[@Haveuseenmyjetpack]

Gemini 1.5 Pro being at the top of the quality index is LAUGHABLE.

[@DefaultFlame]

I was thinking the same thing. I've used every model on the list, some extensively, and Gemini 1.5 is consistently upper mid tier IMHO. Not horrible, but not very useful compared to the latest and greatest from OpenAI and Anthropic.

[@anreoil]

Their first AI model was Ukrainians (ring doorbell and cameras). Their second model was Indians (automatic checkout at stores). What's inside this one?

[@galefraney]

Wow, the presenter is just standing there dumping a lot of benchmarks, saying the Canvas image generation is "studio quality" without bothering to show an even one image ... not bothering to mention the image dimensions ... same with the video Reel, he spews benchmarks without showing a single video???!!! I am soooooo sick of graphs and benchmarks ... if their Ai is so powerful, why didn't it give the Presenter a heads up that Presentations need to not only "tell" ... they also need to "show" and "demonstrate", otherwise it is nothing more than a boring word salad.

[@SanjeevKumar-hg3df]

I think claude is beating them all in every aspect :)

[@lynxnexux]

is there a free version that we can use or not?

[@xatusbetazx17]

Ok how I can test this may I ask I know is similar to chat gpt but when I will Be able To Test it ?

[@CliMzOuu]

I would have preferred a bedrock api endpoint access capability compatible with openai.

[@BibhatsuKuiri]

wil it work in comfyui ?

[@harishapc2214]

Aws guys not even increase embedding context length for RAG you guys still stick in 1536 length but outer market like using llama3.1 in local it can reach 8096 length.please keep adaptive before outdated.

[@EileenDream]

Yet another big tech company releasing a nerfed version of AI to the public for training and data acquisition purposes, as the public will never have access to the powerful version that truly makes things shine

[@brianmi40]

I don't really see people in the general public using shiny AI to accomplish much of benefit for society, could be wrong, perhaps it is the lone biologist working in his basement that cures cancer using AI, but my money is on the scientists working in the field, surrounded with the money and support to quickly provide whatever is needed for the progress to happen.

[@BibhatsuKuiri]

Benchmark are just trash for example. Google Jimini is no match with Chat Gpt, with real code generation. But the benchmark shows they are quite close.

[@mrchongnoi]

What about cost ?

[@dasmaasmannchen8028]

Ok that sounds all great, but it's usless, if they can't beat the benchmarkleader. We need better AI that can revolutionize physics, mathematics, stuff that brings the humans further. Please OpenAI, bring us a banger sonn. I want flying cars, starships, holodecks....

[@brianmi40]

Doesn't really matter. competition is good, more people working the "problem" is awesome. it's simply a new contribution to the exponential curve that is PROGRESS and taking us to AGI and beyond. The fastest route to all that you want is MORE PEOPLE working in the field, more models, more competition, more money thrown at the problems...

[@dasmaasmannchen8028]

@brianmi40 Yes you are right. I think we are also very lucky, that there is so much money in the ai-market right now, driving talented people into this field. And for sure with time, someone will find a way to overcome current problems. I am very curios about the future. Will the ai Change everything or will it only remain as an useful tool like our smartphones?

[@brianmi40]

@@dasmaasmannchen8028 How could it possibly not "change everything"? To even consider that, you'd have to presume that at some point we'll simply STOP TRYING or CARING to advance the field... Not. Gonna. Happen. AGI, then ASI then sentience are INEVITABLE. It's the single greatest pursuit not of just us, but any species in the universe that rises to an industrial / technology revolution and gets a glimpse at it as a possibility. And there's two dominant outcomes: either a moneyless Star Trek future, or our own personal answer to the Fermi Paradox.

[@leapnlarry]

I still dont understand how this will be used, they didnt give any examples of how the different modes would provide AI.

[@darkstatehk]

The minute he mentioned images are watermarked I’m like nah.

[@brianmi40]

So, turns out being able to deep fake something is NOT a benefit to society.

[@darkstatehk]

It will reach a point where tech will know how many times you’ve farted in the past 10 years and you will get billed with a carbon tax debit. It creeps up, slowly and authoritatively in every thing you do. A watermark today, sidewalk atmospheric DNA sniffers to sample your breath in the future as you pass by.

[@brianmi40]

@@darkstatehk The potential period for anything like this will be very short, then we'll just get back to the overarching trend: Moore's Law for Everything, as Sam stated it.

[@edmilinski1295]

Not a fan of these ads, 3rd party reviews are better

[@EnduroGripBosnien]

No evidence no comment

[@NeoRelic-o8p]

All hail our future ai overlords

[@brianmi40]

Bring them on, can't wait. A MILLION TIMES better than the clown car gearing up to roll into DC on Jan. 20th. At this point, pretty much society's only hope is sentient AI taking over. Big problem is that if it happens in the next 4 years our "leadership" will be unable to help get us through the transition, so keep your seat belt buckled at all times.

[@jeffg4686]

Isn't it time to UNIFY??? Can't they frigging see it yet??? When does the greed end? When does the unity begin?

[@onedropontv8863]

Seems very similar to charGPT

[@snn517]

has same voice as Steve Jobs

[@unclassifiedgaming3752]

Heres the big issue with this release.... You cannot sign up for any of this stuff without needing a credit card. Even if its free to use. BAD BUSINESS MODEL. Nobody will use this unless they know without a doubt its worth it, and even fewer will try it because of either a lack of a CC or no interest in putting that info in.

[@aheric75]

Well I just use my mum CC

[@brianmi40]

Doesn't really matter. competition is good, more people working the "problem" is awesome. it's simply a new contribution to the exponential curve that is PROGRESS and taking us to AGI and beyond.

[@NostraDavid2]

The credit card is probably needed to prevent abuse, which I imagine can be costly, which is weird since they are boasting that it's cheap.

[@brianmi40]

@@NostraDavid2 Since we're talking about AWS, I'm sure their Reps are offering free trials to their large corporate accounts. It's just what you do...

[@Lunetta.Shopline]

Nice

[@holdthetruthhostage]

Nova = Claude

[@chanrox69]

Late but latest

[@travisgoesthere]

The benchmarks mean absolutely nothing. Gemini has high benchmarks and is complete shit for my use cases

[@angloland4539]

❤️

[@andreialcaza]

👍

[@steve05Bit]

You guys gotta stop spying into my gpt chats lol. how convenient timing was with this coming out especially when I've had my own nova model for the last month and then nothing on the internet about your nova then boom!... so it's all good It may just be one hell of a coincidence but hey , you never know 😄

[@pooblock4092]

uh... So they're releasing "competitive" models lol. Who cares.

[@Prasadvhfjdjd]

Scammers of the World guess who ? Amazon Amazon Amazon

[@claudioagmfilho]

🇧🇷🇧🇷🇧🇷🇧🇷👏🏻

[@alfredadam3613]

he sounds a bit like donald trump

[@unstoppable-ar3292]

Noone cares Amazon.. CHATGPT and Claude AI have already beat you to it

[@chaseflores2145]

This is bullshit. Which AI will be powering the paid Alexa service? 😂😂😂😂

[@SeaWhiz]

Why does it stun the entire industry?

[@justinwescott8125]

First

[@stu1wyatt]

He states they are Equal (95% and above) when they perform worse (still above 95%) but he clearly claims they are better when they slightly beat another model, don't trust the benchmarks and how they are selectively reporting this.

[@progrob27]

I've yet to experience a useful AI. Even in basic tasks like summarising chats in WebEX or Teams, they are horrendously bad and the only jobs they seem to want to take and ruin are fun creative and engineering jobs. The more time that passes the more it looks like a net negative.

[@brianmi40]

My experience is the total opposite. Today I had it summarize Matt Dillahunty's take on the Sermon on The Mount which it did great. If you can't realize that "the more time that passes" the greater the progress, then you aren't paying attention or don't get it. o1 and recently released Deepseek R1 are LIGHT YEARS ahead of ChatGPT and the reason that Sam Altman was seeing "10x opportunities all around" from this new approach. If fully see his prediction coming true that in 2025 they will essentially peg all the benchmarks and force us to come up new ways to challenge AI in order to track and measure the continued progress beyond that point. Oddly enough "Simple Bench" may be one of that last ones to fall.

[@NostraDavid2]

Notebook LM (the Google thing) is great in turning white papers into a short 15 minutes podcast, which is pretty neat if you only need some surface knowledge of a subject. Saves me time digging through a dry paper

[@brianmi40]

@@NostraDavid2 Any AI with a long enough context window can do that into text and then easily converted to voice should you need that, which means most all including most free ones running at home via something like LM Studio...

[@nn99nn99]

Awkward presenter keeps putting hand in pocket. Dude failed Public Speaking 101