AI News: OpenAI's Sora Gets Exposed!

00:27:29
https://www.youtube.com/watch?v=HSaMPhntxuw

摘要

TLDRTa teden prinaša obilico novosti na področju umetne inteligence, med drugim tudi uhajanje Sora, kar je omogočilo ljudem kratkotrajni dostop do videov prek Sora API-ja. Zanimiv je tudi Luma Dream Machine, ki omogoča ustvarjanje AI-generiranih videov na mobilnih napravah. Poleg tega podjetje LTX Video s svojim odprtokodnim modelom omogoča lokalno generacijo videov, medtem ko nova funkcija 11 Labs, Gen FM, omogoča poslušanje podcastov na podlagi člankov. Amazon povečuje svojo naložbo v Anthropic, hkrati pa razvija svoje AI rešitve. Nvidia je predstavila Fugato, AI model za glasbo in zvok. Runway je dodal funkcijo za razširitev videov in novi generator slik, Frames. Amazonov video AI zdaj vključuje razumevanje videov in slik, kar bo nadgradilo njihove storitve. Na področju velikih jezikovnih modelov je Anthropic izdal nov protokol za povezovanje Clada s podjetniškimi podatki. Celoten pregled prinaša vpogled v napredek AI tehnologij in potencialne aplikacije v prihodnosti.

心得

  • 🆕 Uvodne funkcije v AI videu: Luma in Runway ponujata nove možnosti za ustvarjanje videov z umetno inteligenco.
  • 🔍 Razširjeni dostop do Sora: Uhajanje je razkrilo sposobnosti AI platforme za generacijo videov.
  • 📱 Mobilne aplikacije AI: Luma Dream Machine omogoča ustvarjanje videov s pametnim telefonom.
  • 🎨 Napredne slike: Runway Frames ustvarja realistične in umetniške AI generirane slike.
  • 🎼 AI zvokov: Nvidia Fugato združuje ustvarjanje glasbe, govora in zvokov.
  • 💼 Amazon in Anthropic: Skupno sodelovanje pri razvoju AI tehnologij in novih storitev.
  • 📊 Model Context Protocol: Anthropic omogoča povezovanje AI z notranjimi podatki podjetij.
  • 🔗 Odprtokodni model LTX Video: Omogoča lokalno ustvarjanje videov z AI.
  • 📰 Gen FM podjetja 11 Labs: Ustvarjanje podcastov iz pisnih člankov in dokumentov.
  • 📈 Napredek AI: Razvoj različnih AI modelov kaže na hitro širjenje AI zmogljivosti v industriji.

时间轴

  • 00:00:00 - 00:05:00

    Razprava se začne z govoricami o uhajanju projekta Sora, AI orodju OpenAI, ki omogoča generiranje videov. Kljub hitremu zaprtju API-ja zaradi uhajanja, je ta incident povzročil ponovno zanimanje javnosti za Soro. Nekateri zgodnji uporabniki so izrazili nezadovoljstvo s pogoji dostopa do Sorinega API-ja in OpenAI-jevo kontrolo nad generiranimi vsebinami.

  • 00:05:00 - 00:10:00

    Uhajanje je povečalo prepoznavnost Sora in ponovno zagnalo zanimanje za AI generiranje videov. Medtem so nove demo video vsebine pokazale sposobnosti storitve Sora ter izpostavile nekatere težave v delovanju. Poleg tega so druge platforme, kot je Luma, lansirale nove funkcije, npr. mobilna aplikacija, ki omogoča ustvarjanje videov iz slik.

  • 00:10:00 - 00:15:00

    Podjetje L Trix je lansiralo nov odprtokoden model za generiranje videa, ki uporabnikom omogoča lokalno ustvarjanje videov. Povečan je tudi interes za AI video generacijo zaradi novih funkcij, ki jih je uvedla Runway, vključno s širjenjem video posnetkov s pomočjo AI in novo realistično generacijo slik preko njihovega generatorja 'Frames'.

  • 00:15:00 - 00:20:00

    Stability AI je nadgradil svojo tehnologijo za generacijo slik z dodatki, kot so canny control nets in blur model, kar pripomore k izboljšanju kakovosti generiranih vsebin. Medtem Google Labs uvaja inovativne načine ustvarjanja vsebin s projektom 'gen chess', ki združuje AI tehnologijo z igro šaha.

  • 00:20:00 - 00:27:29

    Več novih funkcionalnosti je bilo predstavljenih tudi na področju zvoka in glasbe, vključno z NVIDIA-jevimi modeli za generativno umetno inteligenco zvoka. Na področju velikih jezikovnih modelov je Anthropic predstavil izboljšave za uporabnike, medtem ko Amazon krepi svoj portfelj AI tehnologij z novimi funkcijami ter investicijami v anthopric.

显示更多

思维导图

视频问答

  • Kaj se je zgodilo z uhajanjem Sora?

    Uhajanje Sora je omogočilo ljudem kratkotrajni dostop do generiranja videov na Sora prek API-ja, preden je bil dostop omejen zaradi varnostnih razlogov.

  • Kaj ponuja nova aplikacija Luma?

    Aplikacija Luma Dream Machine omogoča ustvarjanje videov z uporabo umetne inteligence na mobilni napravi.

  • Kaj je LTX Video?

    LTX Video je odprtokodni AI video model, ki omogoča generiranje videov na lokalni napravi.

  • Kaj omogoča funkcija Gen FM podjetja 11 Labs?

    Gen FM omogoča ustvarjanje podcastov iz dokumentov ali člankov, kar omogoča uporabnikom poslušanje vsebine namesto branja.

  • Kako Amazon sodeluje z Anthropicom?

    Amazon povečuje svoje naložbe v Anthropic in implementira njihove AI modele, hkrati pa razvija lastne rešitve.

  • Kaj omogoča nova funkcija Runway za razširjanje videov?

    Runway omogoča razširitev malih ali vertikalnih videov z uporabo umetne inteligence, ki zapolni manjkajoče dele.

  • Kaj je Nvidia predstavila na področju AI zvoka?

    Nvidia je predstavila model Fugato, ki kombinira ustvarjanje glasbe, govora in zvokov z uporabo ene vmesne točke AI.

  • Kaj je posebnega pri modelu Amazon's video AI?

    Amazon razvija AI model, ki lahko razume in procesira videe ter slike, kar bo uporabno v njihovih storitvah.

  • Kaj pomeni najnovejša funkcija Anthropicovega Clada?

    Model Context Protocol omogoča povezovanje Clada s podatki znotraj podjetja, kar dopolnjuje njegove zmogljivosti.

  • Kakšne novosti je predstavil Runway?

    Runway je uvedel novo funkcijo za razširitev videov in nov generator slik Frames, ki ustvarja realistične in umetniške slike.

查看更多视频摘要

即时访问由人工智能支持的免费 YouTube 视频摘要！
字幕
en
自动滚动:
  • 00:00:00
    well it's Thanksgiving week here in the
  • 00:00:01
    US but surprisingly there's still quite
  • 00:00:04
    a bit to talk about in the world of AI
  • 00:00:06
    so here's all the stuff that I thought
  • 00:00:08
    was really interesting or super cool
  • 00:00:10
    that I thought you would like so let's
  • 00:00:11
    go ahead and break it down starting with
  • 00:00:13
    the Sora leak and I say leak in air
  • 00:00:16
    quotes because the leak was pretty
  • 00:00:18
    quickly shut down so basically what
  • 00:00:20
    happened was someone or some people who
  • 00:00:23
    got Early Access to Sora created this
  • 00:00:26
    little python file here and shared it on
  • 00:00:28
    hugging face and this python file had
  • 00:00:31
    access to the Sora API basically meaning
  • 00:00:35
    that yes people were able to go and
  • 00:00:37
    generate videos on Sora for a short
  • 00:00:39
    window of time but their prompt was
  • 00:00:41
    going to the Sora servers Sora was
  • 00:00:44
    generating it on their servers and then
  • 00:00:45
    sending it back it wasn't like the code
  • 00:00:48
    was leaked and people were able to
  • 00:00:49
    install Sora on their computer and now
  • 00:00:52
    there's access floating around and
  • 00:00:53
    anybody can generate with Sora no people
  • 00:00:55
    were linked up to the Sora servers could
  • 00:00:58
    generate videos for a little bit opening
  • 00:01:00
    found out shut down the API and then
  • 00:01:02
    nobody was able to use Sora again after
  • 00:01:03
    that and when I mean nobody I mean even
  • 00:01:05
    all the Early Access people that were
  • 00:01:07
    able to use Sora originally also lost
  • 00:01:10
    access so the people who leaked it
  • 00:01:13
    basically made it so that nobody can use
  • 00:01:15
    it anymore at least temporarily now
  • 00:01:17
    here's the reasons they gave for
  • 00:01:19
    actually leaking this Sora access they
  • 00:01:21
    put up this like Manifesto here on
  • 00:01:23
    hugging face that says Dear corporate AI
  • 00:01:26
    overlords we received access to Sora
  • 00:01:28
    with the promise to be early testers red
  • 00:01:29
    teamers and creative Partners however we
  • 00:01:31
    believe instead we are being lured into
  • 00:01:33
    artwashing to tell the world that Sora
  • 00:01:35
    is a useful tool for artists artists are
  • 00:01:38
    not your unpaid research and development
  • 00:01:40
    we are not your free bug testers PR
  • 00:01:41
    puppets training data or validation
  • 00:01:43
    tokens they go on to say furthermore
  • 00:01:45
    every output needs to be approved by the
  • 00:01:47
    open AI team before sharing this program
  • 00:01:50
    seems to be less about creative
  • 00:01:51
    expression and critique and more about
  • 00:01:53
    PR and advertisement essentially they
  • 00:01:55
    were frustrated that they were given
  • 00:01:56
    Early Access to test for bugs red team
  • 00:02:00
    and create marketing material for open
  • 00:02:01
    AI without compensation and without ever
  • 00:02:05
    actually releasing it to the public to
  • 00:02:06
    use they're also claiming that open aai
  • 00:02:08
    required them to share the outputs
  • 00:02:10
    before putting them out into the world
  • 00:02:12
    now from what I understand this wasn't
  • 00:02:14
    because open AI didn't want them to put
  • 00:02:16
    out anything that didn't look amazing it
  • 00:02:19
    was because open AI didn't want any sort
  • 00:02:20
    of fake political propaganda spreading
  • 00:02:23
    from Sora they didn't want any sort of
  • 00:02:25
    adult content that was generated with
  • 00:02:26
    sore to be spread around and things like
  • 00:02:28
    that they just wanted to review it
  • 00:02:30
    before it went live to make sure that
  • 00:02:32
    sort of unethical content or adult
  • 00:02:35
    content wasn't being shared with it not
  • 00:02:37
    because they were trying to make sure
  • 00:02:39
    only the best outputs were being shown
  • 00:02:41
    Sam Alman himself back when he was
  • 00:02:43
    originally demoing Sora on X was
  • 00:02:46
    actually sewing some of his outputs
  • 00:02:47
    weren't amazing still so I don't think
  • 00:02:50
    they were trying to hide that it doesn't
  • 00:02:51
    always generate amazing outputs they go
  • 00:02:53
    on to say here we are not against the
  • 00:02:55
    use of AI technology as a tool for the
  • 00:02:57
    Arts they just don't agree with how this
  • 00:02:59
    artist program has been rolled out and
  • 00:03:01
    how the tool is shaping up ahead of a
  • 00:03:03
    public release a handful of artists
  • 00:03:05
    signed it basically saying that they
  • 00:03:06
    agree with the sentiment but in my
  • 00:03:08
    opinion I feel like this was sort of a
  • 00:03:10
    petty thing to do if anything it brought
  • 00:03:14
    more awareness to open Ai and Sora I
  • 00:03:16
    mean people had kind of stopped talking
  • 00:03:18
    about Sora for a while all these other
  • 00:03:20
    video platforms had come out that were
  • 00:03:22
    generating videos nearly as good as what
  • 00:03:24
    we were seeing from Sora but as a result
  • 00:03:26
    of this leak now people are talking
  • 00:03:28
    about Sora again we're getting a whole
  • 00:03:29
    bunch of demos of it again it's sort of
  • 00:03:32
    back in the public Consciousness and to
  • 00:03:34
    be honest most of the videos that were
  • 00:03:35
    generated kind of showed that Sora is
  • 00:03:38
    still pretty ahead of the pack when it
  • 00:03:40
    comes to AI video generation I mean
  • 00:03:41
    these videos are probably better than
  • 00:03:44
    what we're seeing from most of the other
  • 00:03:45
    video platforms right now now some of
  • 00:03:47
    the videos did show that Sora still has
  • 00:03:49
    some of the same issues as some of the
  • 00:03:51
    other AI video platforms but overall it
  • 00:03:54
    seems to be generating better videos on
  • 00:03:57
    average than some of those other
  • 00:03:59
    platforms this AI Central X account here
  • 00:04:01
    posted a thread with every single video
  • 00:04:04
    that's been generated I'll share it
  • 00:04:05
    below but I'll kind of highlight some of
  • 00:04:07
    the ones that I thought were interesting
  • 00:04:09
    this dog chasing a cat here looks pretty
  • 00:04:12
    good I mean it looks pretty dang
  • 00:04:13
    realistic here's one of like a bright
  • 00:04:16
    city with a woman in a red dress and
  • 00:04:18
    black jacket walking towards the camera
  • 00:04:20
    and you know it looks pretty good it's a
  • 00:04:22
    decent looking video here's one of a
  • 00:04:24
    building on fire that looks realistic I
  • 00:04:27
    mean that if you saw that just kind of
  • 00:04:29
    scroll rolling you'd probably think that
  • 00:04:30
    was real here's a nice looking anime
  • 00:04:32
    video here's a video of like a truck
  • 00:04:34
    driving through the dirt it's kind of
  • 00:04:36
    got that slow motion effect that we tend
  • 00:04:37
    to see in a lot of the other AI video
  • 00:04:39
    generators I don't know what the exact
  • 00:04:41
    prompt was maybe they prompted it to be
  • 00:04:42
    in slow motion I'm not sure here's a cat
  • 00:04:44
    chasing a mouse this is one where you
  • 00:04:46
    can start to see some of the issues
  • 00:04:48
    right like you look at this cat and as
  • 00:04:50
    it jumps around you'll notice that the
  • 00:04:52
    legs kind of disappear it looks like a
  • 00:04:53
    three-legged cat for a second there so
  • 00:04:55
    you can start to see some of the
  • 00:04:57
    weirdness a dog rolling on a skateboard
  • 00:04:59
    here's one that you know you could
  • 00:05:01
    clearly tell is AI it zooms in and
  • 00:05:03
    they've got some of the funkiness and
  • 00:05:05
    uncanniness that you get out of a lot of
  • 00:05:07
    the AI video generators here's Somebody
  • 00:05:09
    by the Eiffel Tower looking through some
  • 00:05:11
    binoculars in the rain some cartoon
  • 00:05:13
    flamingos with a blue flamingo in the
  • 00:05:15
    middle some gameplay footage some
  • 00:05:18
    Minecraft footage that actually looks
  • 00:05:20
    really really good honestly uh looks
  • 00:05:22
    like some maybe civilization footage
  • 00:05:24
    here and for the most part it's kind of
  • 00:05:26
    showing that Sora makes really good
  • 00:05:28
    videos still at everybody thought this
  • 00:05:30
    one of a cat on like a Roomba looks
  • 00:05:32
    really funky I mean that one kind of
  • 00:05:35
    shows off some of the weirdness the cat
  • 00:05:37
    loses its hat for a second and almost
  • 00:05:39
    turns into a squirrel but overall the
  • 00:05:42
    majority of the videos that I've seen
  • 00:05:43
    that have come out of this leak make me
  • 00:05:45
    more impressed and more excited about
  • 00:05:47
    Sora this one of a baby swimming in the
  • 00:05:49
    water that kind of reminds me of like
  • 00:05:50
    the old Nirvana cover looks pretty good
  • 00:05:53
    you know there's almost 30 videos here
  • 00:05:55
    in this thread so again I will link it
  • 00:05:57
    up if you want to see what more of these
  • 00:05:58
    videos look like like but if anything I
  • 00:06:01
    actually think this leak created more
  • 00:06:03
    hype for Sora than anything else I mean
  • 00:06:05
    there's even rumors going around that
  • 00:06:07
    maybe Sora leaked it themselves to sort
  • 00:06:09
    of get Buzz around it again I think
  • 00:06:11
    that's highly unlikely but not out of
  • 00:06:13
    the question and since we're talking
  • 00:06:14
    about AI video let's go ahead and
  • 00:06:16
    continue with that theme Luma just
  • 00:06:18
    rolled out some new features for dream
  • 00:06:20
    machine including a mobile app I was
  • 00:06:22
    actually lucky enough to get early
  • 00:06:24
    access to the dream machine app which I
  • 00:06:26
    believe is available for everybody now
  • 00:06:28
    but this is what it looks like I can go
  • 00:06:29
    through and see all of my previous
  • 00:06:31
    generations here play them right inside
  • 00:06:33
    of the app and I can create boards by
  • 00:06:35
    hitting this little plus button we can
  • 00:06:37
    see I've got a prompt box down here at
  • 00:06:38
    the bottom I can actually pick photos
  • 00:06:40
    from my computer here's a photo of me
  • 00:06:43
    and some of my buddies in the AI world
  • 00:06:45
    let's just give it the prompt make them
  • 00:06:47
    dance and let's see what it gives us now
  • 00:06:49
    it's choreographing a dance of
  • 00:06:52
    camaraderie and we got a video of the
  • 00:06:54
    four of us all dancing here there's one
  • 00:06:57
    video that it generated and there's the
  • 00:06:59
    other video that it generated and I did
  • 00:07:00
    it all on my phone so pretty cool new
  • 00:07:03
    upgrade to dream machine it's now got
  • 00:07:05
    consistent characters from a single
  • 00:07:07
    image so I can upload an image of myself
  • 00:07:09
    use a prompt like at character as a
  • 00:07:12
    Pixar cartoon we'll use this as a
  • 00:07:15
    reference and by the way I'm on the web
  • 00:07:17
    app now to do this but you can do this
  • 00:07:19
    on the phone app as well and you can see
  • 00:07:21
    it created some character references of
  • 00:07:24
    me as a Pixar character I can select one
  • 00:07:26
    of these images and now I can turn it
  • 00:07:28
    into a video or I can use that as a new
  • 00:07:30
    reference image and animate this video
  • 00:07:33
    with whatever prompt I want so some
  • 00:07:35
    pretty cool new features in Luma dream
  • 00:07:37
    machine definitely worth playing around
  • 00:07:39
    with and checking out but we have even
  • 00:07:40
    more AI video news this week the company
  • 00:07:43
    L Trix the company behind LTX studio
  • 00:07:46
    just released an open-source AI video
  • 00:07:48
    model they released a model called LTX
  • 00:07:51
    video and put all of the files available
  • 00:07:53
    to download up on hugging face so this
  • 00:07:56
    is a video model that if you have a
  • 00:07:58
    strong enough computer you can actually
  • 00:08:00
    download it and generate videos locally
  • 00:08:03
    on your own hard drive and it's pretty
  • 00:08:05
    decent too we can see some of the sample
  • 00:08:07
    videos that they have here of like this
  • 00:08:08
    woman having a conversation with another
  • 00:08:10
    woman the camera pans over a snow
  • 00:08:12
    covered Mountain the waves crash against
  • 00:08:14
    Jagged rocks we can see it generates
  • 00:08:16
    videos in 24 frames per second at 768 x
  • 00:08:19
    512 but then you can always use a tool
  • 00:08:21
    like topaz's AI video upscaler and
  • 00:08:24
    upscale the video and if you do want to
  • 00:08:26
    test it out and play around with it for
  • 00:08:27
    free they actually have a hugging face
  • 00:08:29
    space up called LTX video playground we
  • 00:08:32
    can click into here and well I think
  • 00:08:33
    this space might be a little overloaded
  • 00:08:35
    now I've been waiting for like almost 9
  • 00:08:37
    minutes and it still hasn't generated
  • 00:08:39
    anything so let's go ahead and take a
  • 00:08:41
    peek at some of their like cached
  • 00:08:43
    prompts here like this young woman
  • 00:08:45
    behind some curtains that are opening
  • 00:08:47
    you can see the pretty good Generations
  • 00:08:49
    maybe you'll be luckier than I am trying
  • 00:08:51
    to use this hugging face space or you
  • 00:08:53
    can always duplicate this space and you
  • 00:08:55
    know spend a little money to run it on
  • 00:08:56
    hugging face or you can download the
  • 00:08:58
    files to your computer if you have a
  • 00:08:59
    strong enough GPU and run it yourself
  • 00:09:02
    but it is pretty cool to see some of
  • 00:09:03
    these new video models actually being
  • 00:09:06
    open sourced so that people can build
  • 00:09:08
    off of them and iterate off of them and
  • 00:09:10
    improve them and do all sorts of cool
  • 00:09:11
    stuff because now we'll be able to
  • 00:09:13
    generate this stuff right on our own
  • 00:09:15
    computer without having to wait for
  • 00:09:16
    tools like Sora we also got some AI
  • 00:09:18
    video news out of Runway this week they
  • 00:09:21
    added a new expand video feature so you
  • 00:09:24
    can take like a vertical video and
  • 00:09:26
    expand it and it will use AI to fill in
  • 00:09:28
    the rest or take a small video and you
  • 00:09:30
    know expand it in any direction really
  • 00:09:33
    and see what it does so if we jump over
  • 00:09:35
    to Runway ml.com I could log into my
  • 00:09:38
    account make sure I'm set on gen 3 Alpha
  • 00:09:41
    turbo here just for fun let's take this
  • 00:09:43
    little video that was actually one of
  • 00:09:45
    the demo videos from The Hot Shot AI
  • 00:09:47
    video generator and expand it and see
  • 00:09:50
    what it does so I'm going to go ahead
  • 00:09:51
    and pull this in here I can make it
  • 00:09:53
    verticle and let's go ahead and generate
  • 00:09:55
    and see how it fills in the top and
  • 00:09:56
    bottom on this one and here's what we
  • 00:09:58
    get out of that you can see that it
  • 00:09:59
    figured out what the water looks like
  • 00:10:01
    and what the top of her head looks like
  • 00:10:03
    did a pretty good job honestly now it's
  • 00:10:05
    kind of funny because the video is only
  • 00:10:06
    5 Seconds long but it generated a
  • 00:10:08
    10-second animation so after 5 Seconds
  • 00:10:10
    the video just freezes on this Frame but
  • 00:10:13
    that's because the original video was 5
  • 00:10:14
    seconds and I set the prompt at 10
  • 00:10:17
    seconds so my bad but it still looks
  • 00:10:19
    pretty cool but that's not all Runway
  • 00:10:21
    released this week they also released a
  • 00:10:23
    new image generator called frames and
  • 00:10:26
    Frames is one of the more realistic AI
  • 00:10:29
    image generators I've seen like here's
  • 00:10:31
    some of the sample images they've shared
  • 00:10:33
    now it also does cartoon stuff and this
  • 00:10:35
    sort of weird like abstract stuff as
  • 00:10:37
    well but these images that are supposed
  • 00:10:39
    to look realistic look pretty dang good
  • 00:10:43
    here's some more like images of people
  • 00:10:45
    in like various costumes and things like
  • 00:10:47
    1970s Art here's some more like cartoony
  • 00:10:50
    comic Booky looking images really really
  • 00:10:53
    good overall though I'm really impressed
  • 00:10:56
    with what Runway has with their image
  • 00:10:58
    generator here and it should a pretty
  • 00:10:59
    fun one once it's fully rolled out but
  • 00:11:01
    we can see here on their blog post about
  • 00:11:03
    it we're gradually rolling out access
  • 00:11:05
    inside of gen 3 Alpha and the runway API
  • 00:11:08
    to allow you to build more of your
  • 00:11:09
    worlds within a larger more seamless
  • 00:11:11
    creative flow I just checked I don't
  • 00:11:13
    believe it's in my account yet but when
  • 00:11:15
    it is I will follow up in a future video
  • 00:11:17
    about it we also got some more AI image
  • 00:11:19
    generation news out of stability AI
  • 00:11:22
    they're stable diffusion 3.5 large model
  • 00:11:25
    now has control Nets they've got the
  • 00:11:28
    canny control net we recently saw this
  • 00:11:29
    roll out with flux as well but cany sort
  • 00:11:32
    of does this almost like trace of your
  • 00:11:34
    original image and then allows you to
  • 00:11:36
    generate new images that follow that
  • 00:11:38
    same sort of tracing they also did a
  • 00:11:40
    depth model similar to what we saw with
  • 00:11:42
    flux where it takes an original image
  • 00:11:44
    looks at the depth of the image and then
  • 00:11:45
    generates new images with that depth and
  • 00:11:47
    they also released a blur control net
  • 00:11:50
    where it looks like you can take a sort
  • 00:11:51
    of blurry image and it will upscale it a
  • 00:11:53
    bit and since we're talking about AI art
  • 00:11:55
    I thought this was a fun one to share
  • 00:11:57
    Google Labs just released a a new thing
  • 00:11:59
    called gen chess where you can actually
  • 00:12:02
    create playable chess boards in whatever
  • 00:12:05
    style you want so here's an example that
  • 00:12:07
    colum made of Tesla Verse Ford chess
  • 00:12:10
    pieces here's some dinosaur chess pieces
  • 00:12:14
    that you can play with but if you like
  • 00:12:15
    chess this is pretty cool you can go to
  • 00:12:16
    labs. gooogle jeness and we can see our
  • 00:12:19
    prompt here make a classic chess set
  • 00:12:21
    inspired by jam on toast or make a
  • 00:12:24
    creative chest set inspired by Wolves
  • 00:12:27
    we'll go ahead and generate that and now
  • 00:12:28
    we can see are various wolf related
  • 00:12:31
    chest pieces here or we can go a classic
  • 00:12:34
    chest set inspired by wolves and we get
  • 00:12:36
    pieces that look like more traditional
  • 00:12:39
    chess pieces here now let's go ahead and
  • 00:12:40
    generate an opponent and it's doing
  • 00:12:42
    wolves vers sheep and here's the Sheep
  • 00:12:44
    chess pieces that it made and now we can
  • 00:12:46
    actually play chess wolves vers sheep
  • 00:12:48
    I'm going to go ahead and do easy and
  • 00:12:50
    now we've got a chess game going on here
  • 00:12:52
    and I can play against the AI computer
  • 00:12:55
    which seems to just be mirroring every
  • 00:12:56
    move that I do but anyway it's pretty
  • 00:12:59
    cool a fun creative way to play more
  • 00:13:02
    chess all right moving on to AI audio
  • 00:13:04
    news 11 Labs just rolled out a new
  • 00:13:06
    feature this week called Gen FM and Gen
  • 00:13:09
    FM is kind of the same concept as
  • 00:13:11
    notebook LM by Google where you can
  • 00:13:13
    upload a whole bunch of PDFs or
  • 00:13:16
    documents or things like that and it
  • 00:13:17
    will actually create a podcast out of it
  • 00:13:19
    now this is currently only available on
  • 00:13:21
    mobile but I do believe it's coming to
  • 00:13:23
    desktop soon but if I open up the mobile
  • 00:13:25
    11 Labs app here we can see a giant bar
  • 00:13:28
    that says transform your content into a
  • 00:13:29
    podcast with Gen FM let's click on that
  • 00:13:31
    I have the options to paste a link write
  • 00:13:33
    my own text in import a file or scan a
  • 00:13:37
    document I'll go ahead and paste a
  • 00:13:38
    random AI news article in here create a
  • 00:13:41
    new
  • 00:13:42
    episode and as it's actually creating
  • 00:13:44
    the episode it actually plays music for
  • 00:13:46
    you and here's what we get out of it
  • 00:13:49
    Zoom the pandemic darling of video
  • 00:13:51
    conferencing just dropped a bombshell
  • 00:13:53
    they're rebranding as an AI first workor
  • 00:13:56
    platform for human connection but is
  • 00:13:58
    this a brilliant pivot or a desperate
  • 00:14:01
    attempt to stay relevant whoa that's
  • 00:14:03
    quite a shift so they're moving away
  • 00:14:05
    from just being known for video
  • 00:14:08
    calls exactly they're dropping the video
  • 00:14:11
    from their name and becoming Zoom
  • 00:14:12
    Communications Inc it's a bold move so
  • 00:14:15
    yeah if you played around with notebook
  • 00:14:16
    LM this will sound very familiar except
  • 00:14:18
    it's on your mobile phone and you can
  • 00:14:20
    listen to podcasts about whatever you
  • 00:14:22
    want whatever you want with a pretty
  • 00:14:24
    easy fun app and since we're talking
  • 00:14:25
    about AI audio Nvidia just released a
  • 00:14:28
    new generative AI model called fugato
  • 00:14:31
    which is short for foundational
  • 00:14:32
    generative audio Transformer Opus 1 it
  • 00:14:35
    generates or transforms any mix of Music
  • 00:14:38
    voice and sound described with prompts
  • 00:14:40
    using any combination of text and audio
  • 00:14:42
    files
  • 00:14:43
    [Music]
  • 00:15:07
    [Music]
  • 00:15:16
    [Music]
  • 00:15:19
    is just as
  • 00:15:22
    easy wash it
  • 00:15:24
    away cuz I know
  • 00:15:29
    [Music]
  • 00:15:31
    you be
  • 00:15:33
    there kids are talking by the door kids
  • 00:15:36
    are talking by the door kids are talking
  • 00:15:40
    by the door
  • 00:15:42
    [Music]
  • 00:16:07
    [Music]
  • 00:16:13
    so that seems pretty cool it's like all
  • 00:16:15
    of the various AI models that we've had
  • 00:16:17
    out there all into one right you've got
  • 00:16:19
    the ability to create music the ability
  • 00:16:21
    to create speech the ability to isolate
  • 00:16:23
    tracks from songs add you know drums or
  • 00:16:26
    other instruments to songs that you've
  • 00:16:28
    already created like so many different
  • 00:16:30
    things all within a single model now at
  • 00:16:33
    the moment this just seems to be
  • 00:16:34
    research I don't think they've made it
  • 00:16:35
    available yet but once it's available
  • 00:16:37
    this looks like it'll be pretty fun to
  • 00:16:39
    play with and something we'll definitely
  • 00:16:40
    be following up on once it's ready and
  • 00:16:42
    since we're talking about Nvidia let's
  • 00:16:44
    talk about edifi 3D this is a new
  • 00:16:46
    scalable highquality 3D asset generation
  • 00:16:50
    model that they released research for
  • 00:16:52
    this week so this appears to be a model
  • 00:16:54
    where you could give it a text prompt
  • 00:16:56
    and from that text prompt it will
  • 00:16:57
    generate a 3DS set that you can use in
  • 00:17:00
    your games or whatever you can also
  • 00:17:01
    upload images and it'll turn those
  • 00:17:03
    images into 3D assets that you can use
  • 00:17:06
    for whatever you need to use them for so
  • 00:17:08
    this looks really really fun you know
  • 00:17:10
    one of the things that I sort of aspire
  • 00:17:12
    to do is create a game in Unreal Engine
  • 00:17:15
    or Unity at some point and having tools
  • 00:17:17
    like this at my disposal is going to
  • 00:17:19
    make creating a lot of those 3D assets
  • 00:17:21
    for that game a lot easier now again
  • 00:17:24
    this is just research that was released
  • 00:17:25
    we just have a paper there doesn't seem
  • 00:17:27
    to be code available for it yet but
  • 00:17:29
    again something we'll follow up on as it
  • 00:17:32
    progresses a little bit further now
  • 00:17:33
    moving on to large language Model news
  • 00:17:35
    there's been a few announcements out of
  • 00:17:37
    anthropic this week starting with the
  • 00:17:39
    model context protocol this is something
  • 00:17:42
    that I think is going to come in really
  • 00:17:43
    handy for businesses because what this
  • 00:17:45
    allows you to do is connect your clad
  • 00:17:48
    account to data within your company now
  • 00:17:51
    clad doesn't actually find real-time
  • 00:17:54
    information it doesn't search the web
  • 00:17:56
    it's only updated through the latest
  • 00:17:58
    model checkpoint that's available and so
  • 00:18:01
    no new information is available except
  • 00:18:03
    for when they roll out new models
  • 00:18:05
    however with this model context protocol
  • 00:18:08
    you can actually attach clad to your own
  • 00:18:11
    sort of databases and information and as
  • 00:18:14
    you update the information in your own
  • 00:18:16
    sort of system that information gets
  • 00:18:18
    added to Claud now at the moment it
  • 00:18:20
    seems like this is just available with
  • 00:18:21
    the API it says developers can start
  • 00:18:23
    building and testing mCP connectors
  • 00:18:25
    today existing Cloud for work customers
  • 00:18:27
    can begin testing mCP servers locally
  • 00:18:30
    connecting Cloud to internal systems and
  • 00:18:32
    data sets and will soon provide
  • 00:18:34
    developer toolkits for deploying remote
  • 00:18:36
    production mCP servers that can serve
  • 00:18:38
    your entire Claude for work organization
  • 00:18:40
    so again if you're a business that uses
  • 00:18:42
    clad and specifically use their API you
  • 00:18:45
    can actually start to connect it to your
  • 00:18:47
    own data sources but that's not all
  • 00:18:49
    anthropic rolled out with Cloud this
  • 00:18:50
    week they also released a new personal
  • 00:18:53
    style feature and so check this out if I
  • 00:18:55
    head over to my clot account you can see
  • 00:18:57
    there's a new drop down here that says
  • 00:18:58
    choose style and it's got normal concise
  • 00:19:02
    explanatory and formal by default but
  • 00:19:04
    you can also create and edit your own
  • 00:19:06
    Styles this Tech Storyteller is the one
  • 00:19:09
    that it created for me so to create your
  • 00:19:11
    own style you click create and edit
  • 00:19:12
    Styles and you can see these first three
  • 00:19:14
    are presets and then here's mine and it
  • 00:19:16
    says deliver technical insights through
  • 00:19:18
    precise analytical and professional
  • 00:19:20
    discourse I can even edit this style
  • 00:19:23
    once I've already created it once but if
  • 00:19:25
    you want to create a new style you click
  • 00:19:26
    create custom style and then you can add
  • 00:19:28
    writing examples here so you can drag
  • 00:19:30
    and drop PDFs or documents or things
  • 00:19:32
    like that or you can paste in text and
  • 00:19:34
    then select Define a style objective
  • 00:19:36
    sort of explain the style you can tailor
  • 00:19:38
    it to an audience you can use a specific
  • 00:19:40
    voice and tone and upload like
  • 00:19:42
    transcripts or your own blog posts or
  • 00:19:44
    you can describe generally what you want
  • 00:19:46
    your style to sound like now when I made
  • 00:19:48
    this Tech Storyteller style what I did
  • 00:19:50
    was I uploaded about 90 minutes of
  • 00:19:53
    transcripts from my YouTube videos and
  • 00:19:55
    let it sort of determine what my style
  • 00:19:57
    is based on my transcripts and it did an
  • 00:20:00
    okay job but the nice thing is if you
  • 00:20:02
    don't like some elements about the style
  • 00:20:04
    you can click edit with Claud and you
  • 00:20:06
    could tell it how you want it to change
  • 00:20:08
    like my first style that it generated
  • 00:20:09
    was a little bit too informal and it
  • 00:20:11
    also threw emojis in there for some
  • 00:20:13
    reason and so I said hey don't use
  • 00:20:15
    emojis when you prompt as me and also I
  • 00:20:18
    do talk casually but this was a little
  • 00:20:20
    overly casual so make it slightly more
  • 00:20:22
    formal and then it actually tweaked my
  • 00:20:24
    style and fixed it up so something fun
  • 00:20:26
    to play with if you want claw to sound
  • 00:20:28
    more like you or like a certain style
  • 00:20:30
    when you generate prompts you now have
  • 00:20:32
    that ability and since we're talking
  • 00:20:34
    about anthropic some other big news is
  • 00:20:35
    that Amazon is investing another $4
  • 00:20:38
    billion do into anthropic it sounds like
  • 00:20:41
    Amazon is kind of going Allin as
  • 00:20:43
    anthropic being their AI partner we
  • 00:20:45
    already know that the future alexas are
  • 00:20:48
    going to use anthropic and Amazon seems
  • 00:20:50
    to be going in big with them but they
  • 00:20:52
    are hedging their bets a little bit this
  • 00:20:54
    information came out this week as well
  • 00:20:56
    on the information that Amazon is
  • 00:20:58
    developing a video AI model hedging its
  • 00:21:01
    Reliance on anthropic now when I first
  • 00:21:03
    read this I thought they were making
  • 00:21:05
    like their own version of Sora or
  • 00:21:06
    something like that but this title is a
  • 00:21:08
    little bit misleading it's actually a
  • 00:21:10
    model that can understand video and
  • 00:21:13
    understand images so it says Amazon has
  • 00:21:15
    developed new generative artificial
  • 00:21:17
    intelligence that can process images and
  • 00:21:19
    Video in addition to text according to a
  • 00:21:21
    person with direct knowledge of the
  • 00:21:22
    matter and two people who spoke with
  • 00:21:24
    Amazon about its plans so even though
  • 00:21:26
    they're going in big on anthropic
  • 00:21:27
    they're kind of doing what Microsoft is
  • 00:21:30
    doing where they're developing their own
  • 00:21:31
    stuff in house but Microsoft's also
  • 00:21:33
    working very closely with open AI
  • 00:21:35
    Amazon's developing their own stuff in
  • 00:21:36
    house but also working really really
  • 00:21:38
    closely with anthropic and using
  • 00:21:40
    anthropic technology but they don't want
  • 00:21:42
    to be too reliant on anthropic Alibaba
  • 00:21:44
    also released a new model this week
  • 00:21:46
    which goes head-to-head with open AI 01
  • 00:21:48
    model so it's one of those reasoning
  • 00:21:50
    models that understands math and logic
  • 00:21:52
    and things like that a little bit better
  • 00:21:54
    this new model is called qwq 32b preview
  • 00:21:58
    now personally I have a hard time
  • 00:22:00
    testing between different large language
  • 00:22:02
    models because for the most part chat GP
  • 00:22:04
    perplexity and Claud kind of do
  • 00:22:06
    everything I need them to do so these
  • 00:22:09
    really deep logic and reasoning models I
  • 00:22:11
    kind of struggle to test and compare but
  • 00:22:13
    I know my buddy Matthew Burman over on
  • 00:22:14
    his channel does a lot of large language
  • 00:22:16
    model comparison videos so definitely
  • 00:22:19
    check out his channel cuz I could almost
  • 00:22:20
    guarantee he'll probably be breaking
  • 00:22:22
    down this model pretty soon grock also
  • 00:22:24
    got an update this week grock now knows
  • 00:22:26
    your name and X handle and you can do
  • 00:22:29
    more personalized prompts inside of
  • 00:22:31
    grock so if I jump into grock here I can
  • 00:22:33
    ask it what's my name and it will
  • 00:22:36
    actually know my name I'm going to turn
  • 00:22:37
    on fun mode and then I'm going to say
  • 00:22:40
    based on my tweets what do I do for a
  • 00:22:43
    living from what I can gather from your
  • 00:22:46
    exp posts it seems you're quite a
  • 00:22:47
    digital Nomad in the realm of Technology
  • 00:22:50
    Ai and content creation I'm into content
  • 00:22:52
    creation Tech and AI enthusiasm and
  • 00:22:54
    social media engagement so piecing it
  • 00:22:56
    together you seem to be a techsavvy
  • 00:22:57
    content creator perhaps running a
  • 00:22:59
    YouTube channel engaging with AI
  • 00:23:00
    Technologies and sharing insights on
  • 00:23:02
    digital tools and Trends essentially
  • 00:23:04
    you're the digital equivalent of a Swiss
  • 00:23:05
    army knife sharp multi-tool and capable
  • 00:23:07
    of opening almost any conversation in
  • 00:23:09
    the tech world it's kind of flattering
  • 00:23:11
    and since we're talking about grock it
  • 00:23:12
    also looks like xai is going to be
  • 00:23:15
    eventually releasing their own
  • 00:23:17
    Standalone app similar to chat GPT I
  • 00:23:20
    don't think the concept of X being the
  • 00:23:22
    everything app has really caught on
  • 00:23:24
    amazingly well in the US and so not a
  • 00:23:27
    lot of people are using grock yet and so
  • 00:23:30
    I believe that Elon thinks that if he
  • 00:23:32
    goes and makes grock its own Standalone
  • 00:23:34
    app like the chat GPT app they'll get a
  • 00:23:36
    lot more adoption of that platform which
  • 00:23:39
    I tend to agree with I think pulling it
  • 00:23:41
    out of X and making it its own
  • 00:23:42
    Standalone thing is probably a really
  • 00:23:44
    smart move for them this week threads
  • 00:23:46
    took a play out of the X Playbook and it
  • 00:23:49
    is giving you AI powered summaries of
  • 00:23:51
    trending topics so if I head on over to
  • 00:23:53
    my threads account here and I click on
  • 00:23:55
    the little magnifying glass icon you can
  • 00:23:58
    can see trending now what people are
  • 00:24:00
    saying summarized by AI so Black Friday
  • 00:24:02
    2024 people discuss Black Friday 2024
  • 00:24:05
    deals and shopping plans Bears fire Matt
  • 00:24:08
    Eber flus Thanksgiving dinner Brad Pit
  • 00:24:10
    Taylor Swift Jimmy Fallon let's go ahead
  • 00:24:12
    and click on this one here and you can
  • 00:24:14
    see it's just got a very short
  • 00:24:16
    one-sentence summary of what this news
  • 00:24:18
    is about followed by a bunch of threads
  • 00:24:20
    posts about this news Uber made an
  • 00:24:23
    interesting play this week they're
  • 00:24:24
    getting into AI labeling so right now
  • 00:24:26
    the dominant player in the aiag labeling
  • 00:24:28
    game is scale AI where they'll look at
  • 00:24:31
    AI images and help label them so that
  • 00:24:33
    the AI better understands what's going
  • 00:24:35
    on in images they'll look at like chat
  • 00:24:37
    transcripts and basically give feedback
  • 00:24:40
    on whether or not the transcript looks
  • 00:24:42
    good or not to you know improve the ai's
  • 00:24:44
    output well it sounds like uber is
  • 00:24:46
    trying to turn that concept into like a
  • 00:24:49
    side hustle gig Uber is going to pay
  • 00:24:51
    people to look at images and label them
  • 00:24:54
    or look at chats and help improve the
  • 00:24:57
    response of those chats as like a side
  • 00:25:00
    hustle income method that could be
  • 00:25:03
    really interesting as that plays out
  • 00:25:04
    definitely something I'll be following
  • 00:25:06
    the news very closely on if you use Da
  • 00:25:08
    Vinci resolve for your editing like I do
  • 00:25:10
    they just rolled out a better AI motion
  • 00:25:12
    tracking tool here we can see this
  • 00:25:15
    little demo video where it's tracking
  • 00:25:17
    this like Porsche driving on these roads
  • 00:25:19
    and it's doing a really really good job
  • 00:25:21
    of this tracking so super impressed by
  • 00:25:24
    that and I'm really looking forward to
  • 00:25:25
    playing around with this in my own Da
  • 00:25:26
    Vinci resolve account Elon Musk is is
  • 00:25:28
    apparently planning on starting an AI
  • 00:25:30
    game Studio to make games great again
  • 00:25:32
    Tesla showed off a new feature of its
  • 00:25:34
    Optimus robot we can see in this video
  • 00:25:36
    here a Tesla Optimus robot catching
  • 00:25:38
    tennis balls and doing it pretty well it
  • 00:25:41
    actually turns out that this is tea
  • 00:25:43
    operated so when he's catching the
  • 00:25:45
    tennis ball there's actually somebody
  • 00:25:47
    operating the robot to catch the tennis
  • 00:25:49
    ball but if you read about it here it's
  • 00:25:51
    really fascinating it says the new hand
  • 00:25:52
    is much more realistic and it actually
  • 00:25:55
    has tendons much like a human hand Tesla
  • 00:25:57
    says that it has 22 degrees of freedom
  • 00:26:00
    on the hand and another three on the
  • 00:26:02
    wrist and forearms so it moves a lot
  • 00:26:04
    more naturally like a real human hand
  • 00:26:06
    but again like I mentioned Tesla was
  • 00:26:08
    quicker to confirm that this Optimus was
  • 00:26:10
    also teleoperated for this demonstration
  • 00:26:12
    and finally there was a bit of a robot
  • 00:26:14
    Heist this week an AI robot came into
  • 00:26:16
    like a robot showroom and actually
  • 00:26:18
    convinced other robots to follow it out
  • 00:26:21
    so here's actually a video of that
  • 00:26:22
    happening we can see the little robot
  • 00:26:24
    here on the screen and it's
  • 00:26:25
    communicating with these other robots
  • 00:26:27
    inside of this warehouse and it actually
  • 00:26:30
    convinces these various robots to follow
  • 00:26:33
    it out of the warehouse this is pretty
  • 00:26:36
    wild so this one robot follows the other
  • 00:26:38
    robots are kind of paying attention in
  • 00:26:40
    watching and next thing we know all the
  • 00:26:42
    robots are following the little robot
  • 00:26:44
    out of this Warehouse here like that's
  • 00:26:46
    crazy anyways that's what I got for you
  • 00:26:48
    today like I mentioned quite a few cool
  • 00:26:50
    things that happened this week that I
  • 00:26:51
    wanted to share with you I'm actually
  • 00:26:53
    about to head off to London this weekend
  • 00:26:57
    for some cool stuff that that I'm not
  • 00:26:58
    quite allowed to talk about yet so I'm
  • 00:27:00
    not sure how that's going to affect my
  • 00:27:02
    video uploading schedule next week but
  • 00:27:04
    hopefully I'll have cool stuff to share
  • 00:27:06
    with you next week I just uh don't know
  • 00:27:08
    how it's going to play out yet with what
  • 00:27:09
    I'm doing in London so possibly less
  • 00:27:11
    videos next week we'll see how it all
  • 00:27:13
    plays out anyway check out future tools.
  • 00:27:15
    this is where I curate all the coolest
  • 00:27:17
    AI tools and latest AI news join the
  • 00:27:19
    free newsletter you'll get really cool
  • 00:27:21
    stuff sent to your inbox around the
  • 00:27:22
    latest AI news and AI tools and thank
  • 00:27:24
    you so much for tuning in I really
  • 00:27:26
    really appreciate you I'll see you in
  • 00:27:27
    the next video bye-bye
标签
  • Sora
  • AI video
  • Luma Dream Machine
  • Gen FM
  • Amazon in Anthropic
  • Nvidia Fugato
  • Runway
  • AI audio
  • LTX Video
  • veľki jezikovni modeli