This New Technology Could Blow Away Gpt 4 And Everything Like It – Just as you’re getting this email, Google has finally released its long-awaited powerful AI, which, in the ongoing tradition of sudden AI name changes, is no longer Bard but Gemini Advanced. I’ve had early access to this LLM for over a month now (remember, I don’t get paid by any AI labs, nor do they see what I write in advance), and I wanted to offer some taster comments.
And yes, I mean taste, not test, notes. In these newsletters, I’ve been careless with spelling – I guess that’s a sign that a normal person and not an AI wrote it – but I’m not making a mistake here. AI testers have their place, but they can also be misleading. AI can be trained on the test questions, either on purpose or by accident, and many of the benchmarks consist of lists of trivia questions or brain teasers that do not reflect real-world use. So I wanted to offer a bit of a subjective/objective mix of opinions on Gemini Advanced, more like a wine sampler giving a rigorous review. I intend to avoid doing a detailed feature set comparison, and focus on the big picture, with lots of examples.
This New Technology Could Blow Away Gpt 4 And Everything Like It
Let me start with the title: Gemini Advanced is clearly a GPT-4 class model. The statistics show it, but so does a month of our unofficial testing. And that’s a big deal because OpenAI’s GPT-4 (the paid version of ChatGPT/Microsoft Copilot) has been the dominant AI for over a year, and no other model has come very close. Before Gemini, we only had one advanced AI model to look at, and it’s hard to draw conclusions with a dataset of one. Now there are two, and we can learn a few things.
People Are Turning To Chatgpt To Troubleshoot Their Tech Problems Now
Some interesting things to note: Gemini searched YouTube, ChatGPT used Bing. Gemini, like ChatGPT, occasionally forgot what it could do and told me it couldn’t create images. Once convinced he could do this, Gemini produced much better images but still lacked precise control – her description didn’t exactly match the image of the shoe she created.
At the same time, Gemini Advanced obviously does not beat GPT-41 in the benchmarks. It’s really good (more rigorous testing is needed to know just how good), but I’d agree with the testing that it’s on par, though it has its own strengths and weaknesses. GPT-4 is much more sophisticated in its use of code and performs a number of difficult verbal tasks better – it writes a better stanza and passes the Apple test. Gemini is better at explaining and does a great job combining images and searching. Both are strange and inconsistent and more delusional than you’d like. I find myself using both Gemini Advanced and GPT-4, depending on the circumstances, as discussed below.
Apple test. Like all word games, this is a challenge for AI, in part because they don’t see words as they do, they work in tokens, which contain one or more parts of words. Still, GPT-4 does well, Gemini doesn’t.
No one has a great definition of sentience, which is fine because LLMs are not sentient by any means; They are software systems designed to create a human-like language. But there’s a strangeness to GPT-4 that isn’t Sentinel, but isn’t like talking to a program either. Weirdness only comes out after you spend enough hours playing with the AI and become excited, or pleased, or both, by its unexpected abilities and apparent intelligence. There was a famous and controversial paper published by Microsoft research shortly after the release of GPT-4, called “Sparks of Artificial General Intelligence”, which tried to put this argument in scientific terms, but ended up simply calling it “Sparks” of Artificial General Intelligence. It is an illusion of a person on the other end of the line, even though there is no one there. GPT-4 is full of ghosts.
Openai O1 Preview Is Finally Out—and It Will Blow Your Mind
Seriously, if you’ve been using the system for a while, I can almost guarantee at least one moment when you get up from your desk, walk around the room, and wonder what’s going on. Here is one example: I asked Twins:
(For context, PbtA refers to Powered by the Apocalypse, an RPG format that’s kind of like Dungeons and Dragons, but more character-driven). Everything you see below is unedited: the actual instructions and first responses from the AI. It’s pretty solid stuff, from the writing to the world building.
There are also some good lines: “The refuge’s plea hangs in the air. It’s not just a summons, it’s an opportunity. It may not give you all the answers, but it’s an opportunity to act, to do
. It’s an escape from the stagnation that has begun to consume you. You don’t dwell long on the decision. One look at your boring crystal set is enough.”
This New Technology Could Blow Away Gpt-4 And Everything Like It
This says something important, I think, which is that GPT-4 “sparks” are not an isolated phenomenon, but may represent an emerging feature of GPT-4 models. When the AI model is big enough, you can get ghosts.
Although still a chatbot, Gemini has a much slicker interface than GPT-4, and is less prone to technical errors, at least in my tests, than ChatGPT. It also has a different “personality” than GPT-4, in its ChatGPT or Copilot incarnations. While GPT-4 is pretty bland (at least since the disappearance of Bing’s personality, Cindy), Gemini is apparently friendlier, nicer, and has a penchant for puns.
Despite these personality differences, it’s amazing how well suited the two very different models are. Complex directives that work in GPT-4 work in Gemini, and vice versa…with some interesting exceptions that align with personality. We have actively experimented with using artificial intelligence for learning and have written articles with suggested suggestions. While updating the guidelines for Gemini (the updated article should be available here soon), we noticed that compared to GPT-4, it is constantly trying to be helpful. In fact, it’s so helpful that it can undermine the purpose of our prompts by trying to help the student, rather than letting them struggle to understand a concept on their own. We had to change our guidelines slightly to reduce this behavior.
Thus, there are differences, but also many similarities. Both systems have safety rails, but they are activated in different ways. Gemini seems more willing to write darker writing than GPT-4, but completely refuses to explain how nuclear bombs work through Taylor Swift’s discography, while GPT-4 is happy to do so.
Harnessing The Power Of Openai Models & Its Business Use Cases
One of the most interesting things about Gemini is how it illuminates a vision of AI as a powerful integrated personal assistant, quite different from Microsoft’s application-specific Copilots or OpenAI’s open GPTs/agents. Microsoft has created tight partners for software like Word and PowerPoint that streamline the user’s workload. OpenAI seems to have an ambitious plan to create autonomous AI agents that can perform tasks without the need for human intervention. But it looks like Google wants to be your assistant.
Previous versions of Bard had impressive ties to the Google ecosystem (Gmail, Google Docs, Google Travel Tools, and more), but were too dumb to use it. They can pull your emails, but either gloss over too many details, or fail to understand the context, in incredibly frustrating ways. I speculated at the time that Google may have built the infrastructure while waiting for a smarter mind to fill it. That seems to be the case.
All the integrations across Google now make a lot more sense. With a smarter mind, in the form of Gemini Advanced, you can start doing some really interesting things that at best seem magical: “Go through my emails, tell me what’s important, and draft responses to each one,” “Look up my next conference and plan a trip I want.” ” But GPT-4 type model is still limited. The AI still identified some details in the email and got confused about its tools on a few occasions (forgetting that it could use Google Maps, etc.). It’s not there yet, but it’s much closer to being an actual assistant, rather than the limited Siris and Alexas we’ve seen before.
This is, in part, why I suspect that Gemini Advanced is the beginning, not the end, of a wave of artificial intelligence development. We can begin to see a world where AI agents act on our behalf. A GPT-4 type model is not powerful enough to run these agents… but we are getting close.
Pau Garcia Posted On Linkedin
This wasn’t a review of Gemini Advanced – we didn’t cover its excellent multi-modal ability to create and view images, or how it integrates search. We didn’t talk about his coding ability, or the fact that he seems to have some code interpreter-like ability to create and run limited Python programs. We also didn’t cover some frustrations, like the fact that he likes to make elaborate plans that he can’t always actually do
Related Post "This New Technology Could Blow Away Gpt 4 And Everything Like It"