Instagram's AI Chatbot Development / OpenAI's Key Model Updates / Meta's Music Generator
Prepare for an exhilarating journey into AI advancements shaking up the digital marketing landscape! In this issue of our newsletter, we are set to explore these captivating subjects:
?? Instagram Reportedly Explores AI Chatbot Development
?? OpenAI Announces Enhanced AI Capabilities with Key Model Updates
?? AI Image of the week: AI generated QR Code Art
?? Meta Releases AI-Powered MusicGen: An Open-Source Music Generator
We hope you find these topics informative and insightful.
Instagram Reportedly Explores AI Chatbot Development
According to recent leaks, Instagram appears to be venturing into the world of AI chatbots. App researcher Alessandro Paluzzi has unveiled information suggesting the development of AI agents capable of answering user queries and providing advice. This new feature would reportedly allow users to select from 30 unique AI personalities to assist in crafting messages.
AI Chatbot's Potential in Social Media Landscape
The introduction of AI chatbots is not unprecedented in the social media landscape. Snapchat's initial foray into this domain with its 'My AI' chatbot, however, was met with mixed reactions due to insufficient age-gating features. Despite the initial setbacks, Snapchat went on to introduce the 'My AI Snaps' feature, which generates AI images for the user. There are speculations about the AI tools Instagram might use for its chatbots, considering its parent company, Meta, already utilizes generative AI in several of its business aspects.
Instagram's Expanding Horizons
Besides exploring AI chatbots, Instagram appears to be considering other ventures as well. Leaked information hints at the development of a stand-alone app named 'Barcelona,' supposedly a competitor to Twitter. This new app is set to integrate seamlessly with Instagram, though the company has refrained from commenting on these rumors. If they prove accurate, Instagram users could see this Twitter-like addition by summer.
OpenAI Announces Enhanced AI Capabilities with Key Model Updates
OpenAI has announced a series of significant updates to its API and generative AI models, GPT-3.5 Turbo and GPT-4. These updates aim to improve the power and performance of AI applications, addressing the evolving complexities and demands of modern work environments and user experiences.
领英推荐
New Functionalities and Cost Adjustments
Among the notable improvements is the introduction of a function calling capability within the Chat Completions API. This new feature enables models to output a JSON object containing arguments for function calls, thereby more efficiently connecting GPT's abilities with external tools and APIs. Developers can use this function to extract structured data, convert natural language into API calls, or create chatbots that answer queries by calling external tools.
Moreover, the updates include a 75% cost reduction on the state-of-the-art embeddings model and a 25% cost reduction on input tokens for GPT-3.5 Turbo. Additionally, the launch of a new 16k context version of GPT-3.5 Turbo enhances its capabilities compared to the standard 4k version.
Implications for Generative AI Applications
The enhancements in OpenAI's models are not only beneficial to developers but also have a far-reaching impact on users. Companies such as Microsoft, Snapchat, Salesforce, and Morgan Stanley have integrated OpenAI's GPT models into their tools, enhancing their products with generative AI. With the newly announced improvements, user experiences across these platforms are poised to improve, providing sophisticated, high-performing AI applications that meet the needs of the modern digital world.
AI Image of the week: AI generated QR Code Art
Meta Releases AI-Powered MusicGen: An Open-Source Music Generator
Meta, known for pushing the boundaries of technology, has recently unveiled MusicGen, an AI-powered tool capable of generating music. Unlike similar offerings from other tech giants, Meta's music generator is open-source, providing a more democratic tool for the innovative manipulation of sound.
Training and Functionality of MusicGen
The functionality of MusicGen is driven by a text description, transforming textual prompts such as “An ’80s driving pop song with heavy drums and synth pads in the background” into approximately 12 seconds of audio. The tool can also be guided using reference audio like an existing song, in which case it aims to mimic both the description and melody. The training of this AI tool involved an extensive library of 20,000 hours of music, which included 10,000 licensed music tracks of high quality and an additional 390,000 instrument-only tracks from media libraries such as ShutterStock and Pond5.
Performance and Accessibility of MusicGen
In terms of performance, MusicGen holds its own in the arena of AI music generation, producing reasonably melodic tunes that can compete with, if not slightly surpass, outputs from Google's AI music generator, MusicLM. However, while the melodies may be pleasant, they don't quite reach the level of sophistication to replace human musicians. The accessibility of the MusicGen tool is a standout feature; Meta has open-sourced pre-trained models of the tool that can be run by anyone with suitable hardware, primarily a GPU with around 16GB of memory. This level of access is a bold move by Meta, expanding the realm of AI music generation to a broader public.