Inferens AI Podcast

Host 3:Introducing the M4 chip by Apple, where Siri's selective hearing becomes a high-definition feature. It's not a bug, it's an 'innovation'!
Host 1:"Ever wondered where AI is heading next? In today's episode of Inferens AI, we're peeling back the curtain on the latest game-changers: from OpenAI's seismic shift in web search to Apple's groundbreaking AI superchip. Why does this matter to you? These innovations are not just reshaping the tech landscape; they're setting the stage for how AI entrepreneurs and researchers like you will navigate the future. Get ready; this journey through AI's newest frontier is about to challenge everything you thought you knew!"
Host 1:Hey, did you catch the latest tech gossip? OpenAI is cooking up something new for ChatGPT!
Host 2:Oh, you mean that fancy feature where it's going to pull info from Wikipedia and blogs, right? And it's going to show pictures too, like a mini Google!
Host 1:Exactly! It's like they're turning ChatGPT into a digital Swiss Army knife.
Host 2:And it's not just any old blogs and Wikipedia. They're even thinking of using Bing for their web search. Who would have thought, eh?
Host 1:You're spot on. And here's a fun fact: some internet detectives spotted a web page search.chatgpt.com, which mysteriously redirected to chatgpt.com for a bit.
Host 2:No way! That's some real Sherlock Holmes stuff right there. But, you know, the platform already has an online search feature. It's just that you gotta pay to play.
Host 1:True, but with this new feature, ChatGPT might just give Perplexity's AI search engine and even Google's Search Generative Experience a run for their money.
Host 2:Well, it's about bloody time someone stepped up to Google. This is going to be one hell of a tech showdown.
Host 1:You said it! And who knows, maybe next time we chat, we'll be using ChatGPT to fact-check our banter!
Host 2:Ha! That would be a riot. But until then, let's keep our eyes peeled on this tech world. It's like a never-ending rollercoaster, isn't it?
Host 1:Couldn't have said it better myself. Until next time, folks, keep your tech game strong!
Host 3:"Ah, the digital age gifts us yet another platform to exchange heated opinions - this time with the added bonus of memes and long-winded blog rants. Because, clearly, what the world needs now is more screens to scream into. Ah, progress, you're doing it wrong!"
Host 1:Hey, you're our resident tech guru, right?
Host 2:Guilty as charged! What's the latest shiny toy that's caught your eye?
Host 1:Well, it's not exactly a toy, but Apple just announced their new M4 chip. And let me tell you, it's like a mini superhero for AI!
Host 2:A superhero, huh? What's its superpower?
Host 1:Imagine this. It's like having a tiny brain in your device that can do things like write captions for your videos in real-time, or even identify objects in your photos and videos. It's capable of doing a mind-boggling thirty-eight trillion operations per second!
Host 2:Whoa, that's like having a tiny Einstein in your pocket!
Host 1:Exactly! And the best part is, it's now available in Apple's latest iPad Pro tablets, with more hardware like Macs expected to follow.
Host 2:That's insane! I can't wait to see what kind of AI features Apple will announce at their June developer conference. I bet they'll be mind-blowing!
Host 1:Oh, I'm sure they will be. After all, as Apple's Tim Millet said, this engine is "more powerful than any neural processing unit in any AI PC today".
Host 2:Well, I'm sold. Time to start saving up for that new iPad Pro! And folks, if you're as excited as I am about this, stay tuned for more updates on this tech marvel!
Host 3:In a world obsessed with the latest tech, we've now got a chip promising to speed up everything, even our existential dread. Because, you know, if there's one thing we needed more efficiency in, it's spiraling into a philosophical crisis at twice the speed. Thanks, Silicon Valley, for turbocharging our angst!
Host 1:my dear, let's take a detour from our usual chit-chat and delve into the world of AI. Have you heard about this French startup, Holistic?
Host 2:Holistic? Can't say I have. But with a name like that, I'm guessing they're into something big?
Host 1:You could say that! They've just bagged a cool two hundred million dollars in the initial phase of their funding round. Now, that's a lot of dough!
Host 2:Holy smokes! That's a lot of zeros. But what's the deal with them? What's so special about Holistic?
Host 1:Well, they're into creating "multi-agent" models where multiple AI systems interact. It's like a digital version of a cocktail party, but with AI systems instead of people.
Host 2:Text Host 2: Sounds like my kind of party! But seriously, that's some next-level stuff. And what's this about them developing an AI model with capabilities others have struggled with?
Host 1:Ah, you've been doing your homework! Yes, they are planning to tackle some of the big challenges in AI, like memory, planning, and tactical decision making. It's like they're trying to create a digital version of a super-smart chess player.
Host 2:Wow, that's some serious ambition. And who's backing them up?
Host 1:Some big names. Accel Partners LP and none other than former Google CEO Eric Schmidt are reportedly among the investors.
Host 2:Text Host 2: Damn, that's impressive! Looks like Holistic is one to watch.
Host 3:Oh great, another AI promising to remember my birthday—as if I needed a digital reminder to mark another year of existential dread.
Host 1:Alright, have you heard the latest Apple gossip? They're diving into the deep end of the pool with AI chips for data centers.
Host 2:Shut the front door! Apple and AI chips? That's like peanut butter and jelly, a match made in Silicon Valley heaven. Spill the beans!
Host 1:Well, according to our friends at the Wall Street Journal, Apple's been as busy as a one-legged man in a butt-kicking contest, developing custom server hardware to run AI programs in data centers. But it's all cloak and dagger stuff, we're not sure if these chips will ever see the light of day.
Host 2:Ooh, a mystery! Apple's always been one for playing their cards close to their chest. They've got a code name for this project, right?
Host 1:You bet! It's called Project ACDC, or Apple Chips in Data Centers. Quite the rock 'n' roll name, don't you think?
Host 2:Ha! ACDC, I love it. But what's the skinny on these chips? Are they going to be used for training AI models?
Host 1:Not quite. The focus is more on AI inference, which is basically the process of using an AI model to make predictions, rather than training the model itself.
Host 2:Ah, I see. So if Apple pulls this off, they'd be joining the big leagues with Amazon, Google, Microsoft, and Meta, right?
Host 1:That's the plan. But not everyone's buying it. The host's Mark Gurman, for instance, is a bit skeptical. He mentioned a previous Apple server chip project that was canned and expressed doubts about Apple developing the chips due to high costs and its focus on on-device AI.
Host 2:Well, well, well, the plot thickens! It's definitely going to be a popcorn moment to see how this pans out. Apple's always full of surprises.
Host 1:Couldn't agree more! Let's keep our eyes peeled on this one. And to our listeners, don't forget to tune in next time for more juicy tech news!
Host 3:"Project ACDC? Seems like their only current is a ‘Highway to Hell’ if they mess this up."
Host 1:You know, in the midst of all this tech jargon, the company is making some serious strides. They're launching a tool to detect images generated by its DALL-E three technology.
Host 3:Fantastic, now we've got AI pretending to be Picasso in a digital beret. Next, it'll be hosting art galleries in the cloud, critiquing its own pixels. Because, clearly, what the world needs is a machine with an existential crisis over abstract art. At least it won't drink all the wine at the opening.
Host 1:Well, in their internal testing, the tool correctly identified ninety-eight percent of DALL-E three generated images. But, just like when you crop your club photos, minor image modifications dropped the detection rates to ninety-five to ninety-seven percent.
Host 1:OpenAI's David Robinson told The WSJ that concerns about upcoming elections are "absolutely driving a bunch of this work.” With the rise in AI-generated content online, there are heightened concerns about election misinformation.
Host 1:Good question! Hue adjustments dropped the detection rate to eighty-two percent. But it's not just about DALL-E 3. The tool could only accurately classify five to ten percent of images produced by other AI models.
Host 1:You're right! A UChicago Harris/AP-NORC Poll found that sixty-three percent of U.S. adults believe tech companies bear "a lot" of responsibility to prevent this type of misinformation.
Host 1:Exactly! You've hit the nail on the head. This image detection classifier, as they call it, "predicts the likelihood that an image was generated" with DALL-E three.
Host 2:Well, let's hope this new tool from OpenAI can help with that. And maybe it can also help me figure out which of my Instagram followers are bots!
Host 1:OpenAI has opened applications for the detector to its first group of testers, including research labs and journalism nonprofits.
Host 2:Yeah, I heard about that. A lot of people think tech companies should do more to prevent this type of misinformation, right?
Host 1:Haha, you might be onto something there! For all our listeners out there, we'll keep you updated on this. Stay tuned!
Host 2:So, it's like a lie detector test for images? That's wild! And how good is this bouncer at spotting the fakes?
Host 2:And what if I go all Picasso and mess with the colors, like adding a funky filter?
Host 2:So, it's not perfect, but it's a start. Who's getting to play around with this?
Host 2:That's some serious stuff. But why now? Why the sudden need for this?
Host 2:No kidding! So, it's like a bouncer for AI-generated images?