Other industries
The AI industry is booming like never before, with new models and innovations popping up all the time. As technology keeps advancing, AI is becoming more powerful and versatile, shaking up different industries. In this article, we’re diving into some of the latest AI releases that are setting new standards and pushing the limits of what these smart systems can do.
Gemma 2
Google AI has released Gemma 2, a new open-source large language model (LLM) designed for researchers and developers. It comes in two sizes (9B and 27B parameters) and offers strong performance, plus, both models run efficiently on a single NVIDIA GPU, reducing deployment costs.
Gemma 2 is fully open-source with a business-friendly license, allowing for both research and commercial applications. It integrates with popular AI frameworks like Hugging Face Transformers and TensorFlow. Additionally, it is trained with robust safety processes to minimize bias and potential risks.
You can access its full potential for free through Google AI Studio, Kaggle, or Colab notebooks. Google Cloud also offers free tier access and credits specifically for research and development purposes. A smaller, more user-friendly 2.6B parameter version is on the horizon to further bridge the gap between lightweight accessibility and powerful performance.
Grok 2
Elon Musk’s xAI is launching Grok 2, the next generation of its language model, this August. According to Musk, this “giant improvement” promises to outperform current models “on all metrics”.
But Grok 2 is just the warm-up. By the end of the year, xAI plans to release Grok 3, aiming to rival (or surpass) the unreleased OpenAI powerhouse, GPT-5. Grok 3 will be a massive undertaking, requiring 100,000 GPUs for training.
ChatGPT 5
ChatGPT 5 is the fifth iteration of OpenAI’s popular large language model. It’s not available yet as it’s still under development, but it is expected to launch sometime in late 2024 or early 2025. There’s no official confirmation on the exact release date yet, but many speculate it’s coming soon.
What capabilities can we likely expect from ChatGPT 5?
One of the most anticipated features of GPT-5 is its enhanced ability to hold natural conversations. Imagine chatbots that can follow the flow of a conversation, grasp the context and hidden meaning, and even respond with appropriate emotions. GPT-5 will be able to adjust its tone and style based on your mood, creating a more empathetic and personalized experience.
Early hints suggest GPT-5 will possess reasoning abilities similar to humans. This includes understanding the context behind your words and the situation you’re in. It will be able to consider various factors and information to provide accurate and relevant responses. For instance, if you’re planning a trip, GPT-5 can analyze your past preferences, current trends, and even weather forecasts to recommend the perfect destination.
GPT-5 is expected to be a multimodal learner, meaning it can process information from various sources beyond just text. This includes images, audio, and potentially even video. With competitors already exploring video processing in AI models, OpenAI is likely to follow suit with GPT-5.
CriticGPT
As AI models get smarter, training them with human feedback (RLHF) becomes trickier. The errors become subtler, making it difficult for humans to pinpoint what’s wrong. OpenAI has introduced a new model built to help with the process.
CriticGPT is a model built on the same foundation as GPT-4, but with a different purpose. It analyzes code generated by another AI model, ChatGPT, and writes critiques highlighting potential inaccuracies. These critiques act as helpful prompts for human trainers, enabling them to identify significantly more issues with the AI’s outputs than they could on their own.
CriticGPT isn’t perfect, but it makes a significant difference. When humans use CriticGPT’s suggestions, they write better critiques compared to working alone. Additionally, CriticGPT helps reduce the number of “false positives” — situations where a harmless output gets flagged as an error.
Claude 3.5 Sonnet
Anthropic recently released Claude 3.5 Sonnet, boasting it as their smartest AI yet. This new model runs twice as fast and costs five times less than its predecessor, Claude 3 Opus. Following the trend of free access, Sonnet 3.5 is available for everyone to play with online and on iOS, with developer options, too.
Source: Anthropic.com
While some might call it the “most intelligent model in the world,” Sonnet 3.5 focuses on different strengths than competitors. It can’t search the web or generate images, but it excels in benchmarks and user interactions. Early users report it as being witty and intelligent. Beyond chatting, Sonnet 3.5 introduces a new feature called “artifacts”, which lets users collaborate on projects alongside the AI.
This release is just part of a larger roll-out. Anthropic plans to release updated versions of Opus (bigger) and Haiku (smaller) later this year, completing the Claude 3.5 family.
Discover Gemini with Revolgy: your AI assistant for Google Workspace
Revolgy offers a 30-day free trial of Gemini to help you explore its full potential within Google Workspace. Gemini is not just a chatbot; it’s an intelligent AI assistant built into your Workspace, compliant with the toughest data security guidelines.
Gemini is a versatile AI assistant that functions as a research analyst, writing coach, and creative partner. It assists with tasks such as crafting compelling emails and documents, generating insightful reports, and brainstorming innovative ideas. Integrated seamlessly with your existing Google Workspace apps, Gemini is ready to use at the click of a button.
Read also: Is ChatGPT Plus still worth it?
