.webp)
Transcript
Speaker 1: This podcast is brought to you by Genetech Solutions. Welcome everyone to the deep dive. Today we're taking a look at, well, what many are calling the next big step in AI. Open AI's GPT-5. People are saying it's more than just an update. Uh, maybe a real game changer for businesses, for creatives, anyone really leaning into AI these days. So, our plan is to unpack what's new, see how it stacks up against the last version, and most importantly, figure out how you can actually use it effectively. Okay, let's just jump right in. There's a lot of buzz, but what really makes GPT-5, different, like fundamentally different from GPT-4?
Speaker 2: Yeah. The big thing I think is what they're calling the reasoning model. You can sort of think of it as giving GPT-5, the ability to well think harder about things. It's not just processing; it's more like deeper analysis. The result is you get these much more, uh, pronounced, more detailed answers, and it's smart about resources too. It kind of flips a switch, you know, uses an older model for simple stuff, but kicks in this advanced reasoning when the prompt gets complicated.
Speaker 1: Uh, automatically. Okay. Automatically deciding. But can the user control that?
Speaker 2: Yes, exactly. And this is really useful. You can actually tell it okay, use heavy machinery now. Just start your prompt with the phrase, now think hard about this. That's the signal. It tells the AI, okay, engage full power for this one.
Speaker 1: Now think hard about this. I like that. Simple, direct. So, if it can think harder, what does that actually look like in practice? What's the real-world impact? Is it just slightly better answers or?
Speaker 2: Oh, it's pretty significant. The output quality, um it's starting to approach what you might expect from say, a PhD student. There's a certain academic depth there. We've seen examples like you ask for a business report from some basic data. GPT-5 doesn't just give you the report, it goes further. It might draft plans for LinkedIn, for Alignable accounts, suggest growth strategies.
Speaker 1: Yeah.
Speaker 2: Things you didn't explicitly ask for. It's going, you know, above and beyond.
Speaker 1: Wow. Okay. Take an initiative like that, that's huge. But you know, it makes me think about the old AI problem, hallucinations. Making things up, has that gotten better?
Speaker 2: That's a really critical point. And yes, there's definite improvement. Hallucinations when the AI just states incorrect facts or, you know, makes up stats. It's a known issue. Open AI is saying GPT-5 is about 45% less likely to do that compared to GPT-4
Speaker 1: 45%. Okay. That's a decent drop.
Speaker 2: It is. And, uh, get this, if you use that reasoning model, that think hard mode, they say it's 80% less likely to hallucinate than GPT 3.0 was.
Speaker 1: 80%. That sounds amazing.
Speaker 2: Yeah.
Speaker 1: But I'm still hearing whispers that even with these big jumps, they still mess up sometimes, maybe like one outta 10 times on common tasks. So what does that actually mean for someone relying on this? Can we trust it really?
Speaker 2: It means you have to be careful. Definitely the reliability is much, much better. It's a far more trustworthy assistant now, but it's not perfect. Not yet. For anything critical, anything high stakes, you absolutely still need a human in the loop. Reviewing, verifying your rule kinda shifts right from creating from scratch to being an expert editor, a final checkpoint. You leverage the AI speed in depth, but the final judgment call that's still gotta be human.
Speaker 1: Right. Okay. So human oversight is still key. Got it. Now thinking beyond just the, uh, the raw intelligence, what else can it do? How does it become more of a, like you said, an assistant in the daily workflow? A Swiss Army knife.
Speaker 2: This is where the Agentic AI part comes in. It's a bit of jargon, but basically means the AI isn't just talking, it can do things, it can interact with other software performed tasks for you.
Speaker 1: Like what kind of tasks?
Speaker 2: Well, the big ones right now are integrations with Gmail and Google Calendar. So, imagine it noticing an upcoming client renewal in your email. Right. It could draft a personalized email to them based on past interactions and then pop a follow-up reminder in your calendar. It can help schedule meetings, automatically reply to certain emails, and it can even try to match your writing style, your tone.
Speaker 1: Whoa. Okay. Matching my tone.
Speaker 2: Yeah.
Speaker 1: That's interesting. So really becomes like a virtual secretary.
Speaker 2: Exactly. It has the potential to really streamline a lot of that admin work. Could even shake up the market for, you know, dedicated scheduling tools or email helpers.
Speaker 1: A virtual secretary.
Speaker 2: Mm-hmm.
Speaker 1: Yeah, I could see how that would free up a ton of time. Okay. But I also heard something about personalities. Like GPT-5 can change how it talks to you.
Speaker 2: Yes. This is another quite unique feature. You can actually choose different tones or personalities for your conversations with it, depending on what you need. There are a few main ones right now. Cynic, robot, listener, and nerd.
Speaker 1: Okay. Tell me about the cynic. Why would I want a cynical AI?
Speaker 2: Mm-hmm. Well, think about it. If you're pitching a new idea, the cynic mode will constantly poke holes in it, predict the worst. It forces you to really defend your idea, find the weaknesses before someone else does. It's a built-in stress testing.
Speaker 1: Huh? Okay. I can see that being useful actually. What about the others? Listener?
Speaker 2: Listener is more like a sounding board. It restricts its own output. Uses phrases like, hmm, I see, or uh-huh but what about Kate? It guides you to the answer rather than just giving it. Great for brainstorming when you wanna think things through yourself
Speaker 1: And robot versus nerd?
Speaker 2: Both give detailed answers. Robot is very fact-based, very dry, almost no emotion or, uh, human-like phrasing. Just the data. Nerd gives you that detail too, but with a bit more enthusiasm. A more human touch may be better for explaining things.
Speaker 1: That's really clever. So, you can pick the interaction style that fits the task, whether it's creative, brainstorming, or you know, just getting the facts great. Makes it much more flexible. Alright, let's zoom out for a second. With all this new power, how does GPT-5 fit into the bigger AI picture? How does it stack up against, say, Google's Gemini or Anthropic’s Claude?
Speaker 2: Good question. GPT-5's big strengths are definitely its core mission learning performance, how adaptable it is, and just how well it works in real world situations. It's setting a high bar, but you know, the competition isn't standing still. Each major player has its own focus, its own niche. Gemini, for instance, is really strong on that deep integration with search, getting real time info. Claude has a heavy focus on safety and transparency, which is critical in some industries.
Speaker 1: Right. Like legal or finance maybe.
Speaker 2: Exactly. And then you have things like Llama, which is huge for the open-source community, focused on accessibility for developers. So, every bot kind of has its specialty, and honestly, we should probably expect them all to make big leaps soon, trying to challenge Open AI in their respective areas.
Speaker 1: So, it's not necessarily about which bot is best overall, but which is best for a specific need.
Speaker 2: Precisely.
Speaker 1: Okay. So bottom line then, for you, the listener, maybe running a business or working on creative projects.
Speaker 2: Hmm.
Speaker 1: What's the big takeaway here? What does GPT-5 really mean for you?
Speaker 2: I think the main thing is that GPTr-5 represents a real evolution. It's moved beyond being just a, you know, a fancy text generator. It's becoming a tool for complex problem solving. It often gives you more than you ask for something polished, something insightful. It feels like it's graduated, you know, from maybe a high school level assistant to more of an academic or industrial peer. So, the conversation really isn't if you should use AI anymore, it's shifted. It's now about how you use these incredibly powerful tools to really supercharge your workflow to innovate,
Speaker 1: Understanding the how.
Speaker 2: Yeah.
Speaker 1: That seems crucial, especially with these more complex capabilities we've talked about.
Speaker 2: Absolutely. And, uh, knowing how to best leverage tools like GPT-5 for your specific goals is key. If you're looking for some help figuring that out, Genetech Solutions actually has a team of senior AI developers. They really know these models inside and out. They can help explore how to integrate this power effectively into your workflow for your business. Might be worth reaching out, scheduling a call with them. You probably won't be disappointed.
Speaker 1: That's great advice. Thank you for walking us through all of that today. It's been a fascinating deep dive into GPT-5. Hopefully you've all got a clear picture now of what this new generation of AI can do and maybe something to think about on your own. What if you could master just one of these new features? Like using that cynic mode to bulletproof your next big idea or setting up the virtual secretary to handle your scheduling. How could that really change your day today? That's all for this deep dive. Thanks for tuning in.
