Top 5 news happening in the AI space:

• Outsmarting GPT-4: The Rise of Fine-Tuned CodeLlama-34B

• AI Legal Modernization Steps into Taiwan's Courtrooms for Drafting Rulings

• Google's Gemini AI Challenges ChatGPT in a Showdown of Titans

• The Future of Multilingual Multimodal Translations with SeamlessM4T

• Google's SynthID Aims to Secure AI Images with Invisible Watermarks

Outsmarting GPT-4: The Rise of Fine-Tuned CodeLlama-34B

In a spirited game of technological one-upmanship, the CodeLlama models have shown they're not to be underestimated. Outperforming GPT-4 on the HumanEval tests, these models have been fine-tuned with a proprietary dataset of over 80k premium programming problems.

The Phind-CodeLlama-34B-v2 takes the crown with a pass@1 score of 73.8%, proving that when innovation meets dedication, new AI champions emerge.

Download it here

AI Legal Modernization Steps into Taiwan's Courtrooms for Drafting Rulings

With the aim of digitizing court processes and decreasing clerks' workloads, the Judicial Yuan has collaborated with Chunghwa Telecom to introduce an AI program for drafting court rulings. This innovative system, trained on data from 1996 to 2021, uses the MT5 model to craft legal documents in Taiwan's format.

Initially, the program will focus on DUI and aiding in fraud cases. With its capacity to draft complete rulings, including verdicts and legal citations, the time-intensive task of draft writing could be significantly reduced for judges.

Google's Gemini AI Challenges ChatGPT in a Showdown of Titans 

Google is stepping up in the AI space with its latest model, Gemini, which some claim significantly outperforms GPT-4. This claim centers on the superior computing power of Gemini. However, OpenAI's CEO, Sam Altman, suggests that these claims may stem from Google's internal marketing.

The competition in the AI domain is intensifying as major tech players such as Google and Meta work on developing advanced AI models. Meanwhile, Google is also making strategic partnerships to enhance its cloud services, despite facing legal issues related to consumer data usage.

The Future of Multilingual Multimodal Translations with SeamlessM4T

SeamlessM4T, the first-of-its-kind multimodal and multilingual AI translation model, is set to revolutionize communication by translating and transcribing speech and text across multiple languages. Supporting nearly 100 languages, the model offers speech recognition, translations between speech and text, and text-to-text translations.

Drawing inspiration from the fictional Babel Fish, the objective is to enable seamless communication between individuals, irrespective of the languages they speak. With prior projects like No Language Left Behind and Universal Speech Translator under its belt, this release signifies a major advancement in universal translation efforts.

Google's SynthID Aims to Secure AI Images with Invisible Watermarks

SynthID, launched by Google's DeepMind, is set to revolutionize AI image detection with its state-of-the-art watermark. Unlike traditional watermarks, SynthID is invisible to human eyes yet can be easily identified by AI detection systems.

This technology addresses the growing deepfake dilemma, ensuring the genuine nature of images, especially critical with imminent elections. Though currently aligned with Google platforms, the vision for SynthID is grand, aiming to become a universal standard for verifying AI-generated content across the internet.

Cool finds around the web

• Taipy - turns Data and AI algorithms into full web applications in no time.

• Comic Factory - Turn prompts into full on comics.

• Skills.ai: AI-powered automation for swift, insightful analytics reports.

• Block Glass - Holographic creation platform for 2D to 3D creations.

• GPT Pilot - AI-driven dev tool, crafts apps with developer oversight.

^{Thanks for tuning into today's edition!}

^{For Daily AI Content 👇}