Two copycat AI tools are under fire 🔥

Ethical Question Marks Surface With 2 New AI Tools

Two new AI tools are making waves this week—but not just for their innovative tech. The rollout of Luma's Dream Machine filmmaking tool and AI-powered search engine Genspark are introducing new conversations about the ethics of borrowing tech, ideas, and inspiration. The question at the end of the day: What makes a copycat?

Here’s what you should know about the new tools…and controversy →

First up: Luma’s Dream Machine promises to revolutionize filmmaking by letting users create high-quality, realistic shots through simple text prompts. It’s trained directly on videos. But the ethical dilemma? It’s about transparency.

• Luma’s trailer for its animated Monster Camp story featured a character that looks an awful lot like Mike Wazowski from Pixar’s Monsters, Inc. 

• This sparked questions about the video sources used to train the model—specifically, whether Dream Machine was trained on copyrighted material without permission.

Luma hasn't disclosed the datasets used for training, and Pixar owner Disney has stayed quiet (for now).

Next: We’ve got Genspark, an AI-powered search engine that creates custom summaries, or "Sparkpages," in response to user queries. 

• For example: If you type in “best baby formula for newborns,” Genspark gives you a comprehensive overview from various web sources.

• Similar to fellow AI search engine Perplexity, Genspark aims to deliver high-quality results by using multiple specialized AI models.

The ethical dilemma here? Like Arc browser’s Arc Search and Google’s AI Overviews, Genspark’s use of AI to repurpose existing content without acknowledgment or approval has been criticized for not always giving proper credit to original content creators.

NVIDIA Could Take LLM Performance to a New Level

NVIDIA's new tool could revolutionize LLM performance. Let’s break it down.

The who and what: AI and graphics processing powerhouse NVIDIA has announced Nemotron-4 340B, a family of open models designed to generate synthetic data for training large language models (LLMs) to be more accurate.

• These models are set to transform industries like healthcare, finance, manufacturing, and retail by offering a scalable solution for generating high-quality training data.

• The models include base, instruct, and reward versions that work together to create synthetic data.

The why: High-quality datasets are often expensive and hard to come by. Nemotron-4 340B tackles this issue by providing a free, scalable way to generate synthetic data, making it easier for companies to innovate without breaking the bank on data acquisition.

The how: Using NVIDIA’s NeMo framework, developers can customize the Nemotron-4 340B Base model for specific use cases, enhancing data quality and model performance.

• These models are optimized for inference with NVIDIA TensorRT-LLM, which the company suggests can increase efficient performance at scale. 

• This setup allows developers to fine-tune their models pretty effortlessly.

Be the Next AI Innovator

Don’t just watch the AI revolution—be part of it. Brilliant makes it easy to master the concepts powering AI—with thousands of quick, interactive lessons you can do on your phone.

Here’s what to expect from Brilliant:

• Master AI concepts through fun, bite-sized lessons

• Learn hands-on use cases and get real-time feedback

• Stay up-to-speed on cutting-edge AI topics like LLMs and neural networks

• Build a daily learning habit with fun, gamified features like Streaks and Leagues

Join the 10 million people on Brilliant and start your 30-day free trial today. Plus, Future Tools readers get a special 20% off a premium annual subscription.

Your New Customer Support AI

This week, Decagon stepped out of stealth with a shake-up to the AI-powered customer support scene. 

The platform promises to mirror human capabilities and has already snagged big-name clients like Eventbrite, Bilt, Webflow, Rippling, and Substack. But enterprise solutions that promise human-like AI isn’t a new pitch.

So how is Decagon cutting through the noise to land these deals? Let’s get into it.

What sets Decagon apart: Unlike traditional chatbots, Decagon’s AI agents provide contextualized, conversational interactions throughout the entire customer support lifecycle. These agents can reason through complex business logic, take actions, improve from feedback, tag conversations, analyze trends, and even write new articles.

How it works: Decagon uses a mix of fine-tuned and third-party models to train its AI…like an organization’s knowledge bases and historical customer conversations, which give Decagon’s agents personalized responses.

• Google DeepMind debuts an AI that turns video pixels and text prompts into soundtracks.

• Adobe Acrobat reinvents document creation with multi-format, AI-powered features.

• OpenVLA launches as an open-source model enhancing generalist robotics capabilities.

• Snap unveils advanced AI to push augmented reality to the next level.

• Microsoft releases Florence-2, a versatile AI model for diverse vision tasks.

• OpenAI’s former chief scientist launches a new AI venture.