Here are this week's most-read AI stories. To access more news and insights, stay updated with our email newsletter.
1. Google Offers Devs Browser-based AI Workspace in Project IDX
Google has unveiled an experimental new web-based developer workspace platform called Project IDX.
IDX gives developers access to a multiplatform app development workflow accessible from a browser.
The developer platform supports various frameworks, including Next.js, React and Angular – with Python and Go support “coming soon.”
Project IDX users can preview their apps as if users would see them – with Google planning to add support for built-in multi-browser web previews, Android emulators and iOS simulators.
To access Project IDX, you need to join the waitlist for the limited preview.
--- Like what you're reading? For more stories like this on AI, sign up for our email newsletter to stay updated!
2. Vicuna LLM Commercially Available, New v1.5 Update Improves Context Length
The popular language model Vicuna has been given an upgrade and for the first time you can use it to power your commercial applications.
Vicuna v1.5 is a fine-tuned version of Meta's Llama 2 and now comes with an extended context window so that it can handle lengthier documents.
The original Vicuna has a context length of under 5K. Vicuna v1.5 now has 4K and 16K context lengths.
The new model also boasts a far greater accuracy at natural language processing tasks and blew the original models out of the water on the MT-Bench benchmark, which assesses a model's dialogue capabilities.
Each version of Vicuna comes separately – meaning you can choose to download your preferred parameter size with either 4K or 16K context length. Users can now access all weights and sizes via the LMSYS Org page on Hugging Face.
3. EY Survey: Top Management ‘Slow to Embrace’ Emerging Tech
Nine out of 10 U.S. workers say adopting emerging tech will help their companies but top management is slow to adopt them, according to EY’s Emerging Tech at Work Survey.
Around 1,000 employees were surveyed on their views of AI and machine learning and other emerging technologies.
Nine out of 10 employees said that adopting these technologies will benefit their companies.
However, 59% also said top management is “slow to embrace” these potentially “game-changing technologies.”
With slower adoption, these technologies are already out of date by the time their firms adopt them in the workplace, said 52% of respondents.
4. Meet Humpback: Meta’s New AI Model That’s a Whale of an Upgrade of Llama
Meta has come up with a new way to supercharge their AI models – unveiling Humpback, a supped-up version of Llama 2.
Humpback was built using instruction back translation – a new method from Meta that allows an AI model to self-curate the data it uses to generate outputs.
Humpback can self-curate its own data predicting its quality to improve results.
Meta’s resulting whale-themed model is powerful – it outperforms the likes of Claude, Vicuna and Claude at following instructions. Humpback was among the top-performing models on the AlpacaEval test for instruction following models.
Meta’s researchers said they’d look at scaling the method that built Humpback in future tests.
5. ChatGPT’s ‘Custom Instructions’ now available to free users
OpenAI has opened up access to ChatGPT’s Custom Instructions - which lets users customize the chatbot’s responses or output.
Non-premium users except those in the EU and U.K. can now access the feature. Previously, only ChatGPT Plus subscribers could access Custom Instructions.
Custom Instructions lets users set how they want ChatGPT to respond – with the feature also applicable when using plugins.
About the Author
You May Also Like