Select Page

Original & Concise Bullet Point Briefs

GPT-4 BEATEN, Deepminds New AI, Googles Secret AI,

Google and AI AMD Push Boundaries of Artificial Intelligence

  • Google Deepmind released Tappy, a project to independently track any point in a video
  • Google Brain Team works on unprecedented photo realism and deep level of language understanding called Imogen
  • China’s Baidu claims that their Ernie Bot has beaten Chat GPT on key tests in the race for AI development
  • Google announced Virtual Try-On feature, allowing users to see how clothes fit on different body shapes and sizes
  • Google also announced updates to its search generative experience, giving AI generated responses rather than single search queries
  • AI AMD released a new AI chip specialized.

Tech Giants Compete in AI Race: AMDs New Chip, Nvidias h100 GPUs, and UN Summit with Robots

  • AMD released a chip designed to be used for generative AI
  • Nvidia has the Lions Share of the graphics card market, but AMD’s new chip may attract interest from big cloud providers such as Amazon or Microsoft
  • Generative AI has allowed for faster drug discovery to occur in a fraction of the time and cost compared to traditional methods
  • Nvidia’s h100 GPUs have set a new standard for generative AI
  • MetaAI demonstrated a robot that could take natural language commands and avoid collisions with humans
  • The UN is set to host a summit on AI where tech gurus will discuss with 50 plus robots
  • OpenAI and Microsoft have been sued for alleged Chachi BT privacy violations.

Microsoft and Open AI Sued Over Unpaid Image Creators, Drag & GPT-4 Push Language Model Limits

  • Open AI and Microsoft are being sued for not paying the creators of images used in their AI platform
  • Drag is a software which users are able to manipulate images as if they were 3D models
  • Extending context window of large language models allows them to process larger volumes of texts
  • GPT-4 already has an implemented 32,000 context window length.

Original & Concise Bullet Point Briefs

With VidCatter’s AI technology, you can get original briefs in easy-to-read bullet points within seconds. Our platform is also highly customizable, making it perfect for students, executives, and anyone who needs to extract important information from video or audio content quickly.

  • Scroll through to check it out for yourself!
  • Original summaries that highlight the key points of your content
  • Customizable to fit your specific needs
  • AI-powered technology that ensures accuracy and comprehensiveness
  • Scroll through to check it out for yourself!
  • Original summaries that highlight the key points of your content
  • Customizable to fit your specific needs
  • AI-powered technology that ensures accuracy and comprehensiveness

Unlock the Power of Efficiency: Get Briefed, Don’t Skim or Watch!

Experience the power of instant video insights with VidCatter! Don’t waste valuable time watching lengthy videos. Our AI-powered platform generates concise summaries that let you read, not watch. Stay informed, save time, and extract key information effortlessly.

let's take a look at some of the thingsthat happened this week in artificialintelligence that you may haveunfortunately missed lots of these arereally interesting so watch until theend because trust me when I say you'llbe Mighty surprised by what you seecoming in at number one is of courseGoogle's deep mind they actually madethis really cool project called tapiressentially what this project does isit's able to independently track anypoint in a video now you might bewondering why is this really cool wellif we compare it to other examples thatwe've seen before it just isn't aseffective you see what tap here is ableto do is get a point on an object andkeep it there for a very long time andif you're wondering about theapplications this just has many and manydifferent applications from enhancingall video editing software to being ableto track things independently of what wesee it's definitely going to change alot of Industries but of course as youknow this is still in its early stagesso with Google deepmind we know knowthat they constantly release new thingsand I've got to be honest tappy is oneof those things that I will think isgoing to be great for the future then wegot this state-of-the-art project fromGoogle Now what you're looking at isImogen unprecedented photo realism and adeep level of language understanding sothis is from Google researchers brainteam now essentially why this is so coolis that this wasn't just a researchpaper Google had been working on thisfor quite some time and finally theymanaged to release this so what you cansee here is a dog looking into themirror seeing a cat a dragon fruitwearing a karate seat belt in the snow Aporanium sitting on the throne there aremany different things that you can thinkabout now essentially if you'rewondering what exactly is this is thislike mid-journey well yes it's likemid-journey but what's cool about thisis that Imogen isn't just an imageeditor they also have image and videoand an image in Video Edit now image andvideo as you can see right here isn't asgood as many other things but it isstill in its early stages where you'reable to generate great video from textnow what was also cool about thisrelease by Google was they actuallyreleased this to the public so from thisvideo that you're able to see you cansee that you can write a prompt in thetext box below and upload an image andthen of course you can edit it it'sactually really really similar toadobe's new firefight thing so I wouldsay once again Google are showing thisthat literally every week they'recommitted to releasing new andInnovative AI products that we are goingto be using and finally finally we didget this one released so this is ofcourse image vertex on AI and I'll leavea link to this one in the descriptionthen of course this is the title of thevideo apparently China's Baidu claimsits Ernie Bot Beats chat GPT on keytests as artificial intelligence raceheats up so Baidu said on Tuesday theartificial intelligence model calledErnie 3.5outperformed openai's chat GPT and gpt4in several key areas the ChineseSearcher revealed that Ernie in Marchand has since publicly been testing itin China the chatbot is based on baidu'sfoundational AI model called Ernie thefocus on baidu's advancements underscorethe intense competition taking place inthe area of generative AI withtechnology Giants in the United Statesand China rapidly advancing developmentsso this article basically states thatgpt4 was beaten a large language modelcreated by Chinese developers now ofcourse an important distinction is tomake that Ernie 3.5 developed by Chinadid beat gpt4 in Chinese language testsaccording to the science journal so thisgoes to show that although these aresome small headways that certainindividual AI companies are making it isimportant to note that rapid artificialintelligence development is going tocontinue regardless of whateverjurisdiction that we are in of coursethere are some people calling for theSlowdown of gpt4 and of course theSlowdown of the development of gpt5 butwe can see here across the globeeverywhere companies individualscorporations are rushing to make theirown AI chatbots which they of coursetrying to surpass tpt4 with as is rightnow that is the very best Benchmark soit will be interesting to see maybe if amonth from now two months from now threemonths from now gpt4 does completely getbeta then of course we had Googleannounce their virtual try on featurewhich is going to revolutionize onlineclothes shopping essentially what itdoes it allows users to see how closelook on Real Models with different bodyshapes and sizes I think this is a greataddition to the workspace because thisis going to allow individuals who havemaybe a unique body to be able to seeexactly how that piece of clothing fitsonto their body it was a reallyintriguing research paper but we knowstuff like this is going to beimplemented because we've seen smallthings like this before it's definitelyreally really cool and I think rapid AIadvancements like this are ones thatdefinitely make the world easier to useand a better place now what Google alsoincluded was they also announced thatthey're updating their search generativeexperience this was something that tookEveryone by storm but at the same timewe weren't that surprised because we didnotice that they were talking about thisin their Google I O event so of courseessentially what it does is it gives anAI generated response rather than asingle search query so you know how onGoogle sometimes you search somethingand you really can't find what you'relooking for with sge what they do isthey tailor it specifically to you toall your cookies to all your preferencesand then you're able to talk with thechatbot and find exactly what you'relooking for I find that these kind ofIntegrations into the web are definitelygoing to be much more effective andsince people do use Google anyways overBing I think that that is going toremain the user Choice the only thing iswhen will they release this to thepublic as you know Google is slowlytrying to rush out these products but atthe same time they're quite scaredbecause their AI hasn't proven to be toogood just yet then of course we had acompany called AI AMD released a new AIchip specialized for generative AI now Ihave to be honest with you guys if youdon't know who AMD is then maybe you'renot someone who's that much intocomputers but for those of you who areyou likely know who AMD are if you'veheard of Nvidia you'll likely have heardof AMD they're essentially directcompetitors with Nvidia taking theLion's Share of the graphics card marketnow AMD what they've done is they'vereleased this card to the supply forgenerative AI related products andapparently they described their m1300xchip as the world's most advancedaccelerator for generative Ai andapparently it's expected to attractinterest from big cloud providers suchas Amazon or Microsoft but they haven'tactually specified who's going to beusing it and it will be interestingbecause if AMD is able to create a superchip that is far superior than nvidia'sthen it's likely going to push Nvidia topush the boundaries even further and tobe honest with you guys Nvidia does seemto be doing things at a increasinglyfaster rate with many differentadvancements in AI that we're going toget then of course we had somegroundbreaking news where we had quickercures how in Silicon medicine usesgenerative AI to accelerate drugDiscovery so essentially in silico usedgenerative AI for each step of thepre-clinical drug Discovery process toidentify a molecule that a drug compoundcould Target to generate novel drugcandidate gauge how well thesecandidates would bind with the Targetand even predict the outcome of clinicaltrials and doing this using traditionalmethod would have cost more than 400million dollars and taken up to sixyears but with generative AI in silicoaccomplished them for one tenth of thecost and one-third of the time reachingthe first phase of clinical trials andjust two and a half years afterbeginning the project so like we statedbefore although AI might not just makeone huge jump it's every smallincremental jump that advanced societyas a whole in fact when we also look atanother Nvidia blog post we can see thattheir recent h100 gpus have set a newstandard for generative AI in the debutmachine learning performance Benchmarkso this is absolutely incredible we canseem that with these advancements ingenerative Ai and of course inprocessing power and in chip speed weknow that large language models and allsorts of artificial intelligenceprograms including the training of thesemodels including the retraining of thesemodels is going to get a lot fasterespecially with nvidia's race todominate the entire AI industry we couldsee rapid improvements in artificialintelligence so it's important to notethat although these changes aregroundbreaking in terms of a primary uselevel if we can get the hardware toincrease the speed at which we can getthings done we could see an increasedrate of actual research and Discoveryfrom drug Discovery as we just talkedabout to other machine learning researchthen we had something that was reallyinteresting sting in terms of Roboticsand artificial intelligence as you knowAI is eventually going to be involved inrobots but so what we had here wassomething released by meta AI so if youdon't know of course Mark Zuckerberg hasa company and they do have a very robustAI division in that division they'reconstantly working on new things andthis was something really cool so theydemonstrate the ability to use your ownlanguage in order to talk to your robotand to get it to do very effectivethings so here you can see the user hassaid bring me the box of chocolates thecereal box the pill bottle and put themon the bedroom table I mean for theaverage person this might not pose anygreat application but can you imaginethis in a care home in a home with aperson who has specific disabilities oris unable to reach certain items thiscould most certainly provide a veryuseful application for those tasks andespecially where sometimes you might notbe able to type just being able to useyour voice to a robot to be able to getyou something afford to be able tounderstand you is going to be veryeffective now what's also cool aboutthis is that in terms of the robustnessthe navigation policy is able to take analternative route to avoid collisionwith humans in addition the pick policyis able to change the grasping locationif the object moves of the base is notclose enough to the receptacle andessentially what that means is that thiskind of robot isn't just going to fallover if there is going to be some milddisturbance as you know life has itsdisturbances sometimes we might fallover objects might drop and these robotsare able to adapt to their situationeven if these things occur so take alook at some of these tasks because it'svery likely that in the future maybe youmight not see these robots in your homebut maybe through our cities or maybe incertain workplaces robots that takenatural language commands are going tobe very common especially once we domanage to get the cost of these robotscompletely down it's going to beinteresting to see everywhere that theycould be then something from singularitynext week the United Nations is set tohost a summit in Geneva where Tech gurusare going to be discussing AI alongside50 plus robots this is going to beextremely interesting as we are going tosee The Wider conversation surroundingartificial intelligence it's importantto know that viewers like you who dowatch this channel are definitely moreclued in on artificial intelligence thansome of our world leaders and althoughthat might seem like an outlandishstatement as you can see from congress'sgrilling of some Altman this couldn't befurther from the truth artificialintelligence is moving very very quicklyand sometimes people just don't havetime to catch up so it will beinteresting to see where the world'smind is on artificial intelligenceespecially with these robots as wellthen we had something about open Ai andthis one was also quite controversialyou see open Ai and Microsoft wereallegedly sued for three billion dollarsover alleged Chachi BT privacyviolations the lawsuit claimed the openAI secretly scraped 300 billion wordsfrom the internet the lawsuit which wasfiled on June 28th in federal court inSan Francisco includes 16 Anonymousplaintiffs claimed that openai secretlyscraped 300 billion words from theinternet without registering as a databroker or obtaining the proper consentnow this is something that is going tocontinually happen I mean Adobe seemslike the only company who's trying toactually protect people's data when it'sactually taking their content from theinternet we know that mid-journeyearlier this year did actually have asituation where people are trying to suethem because of course as you knowmid-journey scraped millions of imagesfrom the internet in order to buildtheir AI platform and they didn't payanything to the creators of the originalAI art that they stole it from so itwill be interesting to see how thisplays out in the court of law but I canassure you that open Ai and Microsofthave billions of dollars to fight thislawsuit and I will be wondering if thisis going to go to to a settlement orthis will change the way that data lawsare going to be created and how certainlarge language models are going to becreated as you know currently manypeople just take the data from theinternet as is and is it going to be nowa requirement that when training a largelanguage model you need to verifylegally what kind of data it was trainedon and where you got that data from thatcould have severe implications becauseif you have to go through an extendedprocess where you need to verify all thedata sources it could mean a hugeslowdown because data collection isquite hard although it could also meanthat data collection improves as youknow when looking at large languagemodels data is the number one bottleneckin language model performance so it willdefinitely be interesting to see howthis plays out then of course we had thevery interesting software that isactually very unique in terms of what itdoes this software is called drag againessentially it's a software where youcan manipulate any image by dragging apoint on an image and moving it as if itkind of was a 3D model I mean this kindof technology is only possible with AIand I'm wondering what the trueapplications of this will be maybePhotoshop May integrate this into itssoftware but it is still a little bitbuggy but it does have quite a lot ofapplications for just manipulating animage in a way that we never thought waspossible so Dragon I gotta be honest itlooks really weird but at the same timeit is very very and whilst you're takinga look at some of the examples let meknow if you're going to be trying outthis Dragon because as you know lots ofAI papers do get released a week andreally often do we get the ability totry these out so with this it's actuallyvery effective that we can now actuallytry this out and use it as well as wewant to then we had another researchpaper extending the context window oflarge language models this work doneextends the context window of largelanguage models like Lama up to 32 000tokens essentially what you can do isyou can have llama remember the last 32thousand tokens in your messages whichessentially means it gets more effectiveat processing larger volumes of text andthis is what we want to see because asyou know one of the main bottlenecksfrom the earlier versions of chat GPTand the early large language models isthat they simply cannot remember whatwas said earlier on in the conversationwhich leads to a very limited functionin terms of certain applications withthis paper and with others that we'veseen before this is going to prove veryeffective at allowing us to do much muchmore from analyzing books to analyzingbank statements to analyzing poetry thisis going to be something that we couldall use very well and of course asyou're seeing this is going to be widelyused with gpt4 already having its 32 000context window length alreadyimplemented