Author: IBL News

IBL News | New York

Databricks released yesterday an improved version of its open-source, free-to-commercialize, large language model (LLM) with 12 billion parameters, called Dolly 2.0.

Based on the EleutherAI pythia model family, Dolly 2.0 has been “fine-tuned exclusively on a new, high-quality human-generated instruction following dataset, crowdsourced among Databricks 5,000 employees during March and April of 2023.,” according to the company.

“We are open-sourcing the entirety of Dolly 2.0, including the training code, the dataset, and the model weights, all suitable for commercial use. This means that any organization can create, own, and customize powerful LLMs that can talk to people, without paying for API access or sharing data with third parties.”

Under the licensing terms (Creative Commons Attribution-ShareAlike 3.0 Unported License), anyone can use, modify, or extend this dataset for any purpose, including commercial applications.

IBL News | New York

The top 10 AI mobile apps have already pulled in over $14 million to date, according to analytics provider data.ai.

These apps are mostly dubious ChatGPT apps claiming to be associated with OpenAI in order to charge hefty subscription prices for accessing ChatGPT — a service that’s been free via the web.

This phenomenon is an indication of how much ground OpenAI has ceded in the mobile app market by not having its own official mobile app available.

For the most part, none of this group of apps are trying to establish their own brand and identity. Instead, they’re keyword-stuffing their apps’ titles to match the search terms of people who are looking for ChatGPT or similar, which are mostly “AI,” “Chat,” or “Chatbot” and “Assistant” will help drive.

The majority of the top 10 apps saw little consumer spending before ChatGTP’s addition, in December 2022.

The only exception is the app Pixelcut AI Photo Editor, a “magic writer” copywriter tool that uses GPT. Its creator AI Photo Editor generated $19.8 million in 2022.

In addition to Pixelcut, the other nine earning apps are: Genie – AI Chatbot; AI Chat – Chatbot AI Assistant; AI Chatbot – Open Chat Writer; Apo – AI Personal Assistant; Chat AI Bot – Writing Assistant; ChatOn – AI Chatbot Assistant; AI Chat – Ask Anything; Chat AI – Ask Anything; and GoatChat.

Of these, Genie has generated the most revenue this year, with $3.2 million in global consumer spending so far in 2023.

IBL News | New York

AI PowerPoint-style storytelling startup Tome recently raised $43 million in Series B funding.

Lightspeed Venture Partners led the round while individual investors included Coatue, Greylock, Stability.ai CEO Emad Mostaque, and former Google CEO Eric Schmidt.

While still in a pre-revenue stage, the company is now valued at $300 million post-investment, up from $175 million at its 2021 Series A round, according to an article in Forbes.

San Francisco-based Tome says it’s the fastest productivity software maker to ever reach one million users since its September release.

Tome’s AI software rethinks the PowerPoint system organizing almost real-time presentations in PowerPoint or Google Slides, with slides organized by a table of contents.

“We set out to build a company that can help anyone tell a compelling story,” the Cofounders of the company, Keith Peiris and Henri Liriani, said. “Storytelling is the elementary building block of productivity for humanity, from cave drawings to stories around the fire to PowerPoint.”

Tome uses large-language models to make queries behind the scenes to fulfill a user’s presentation prompt. First, its software generates an output framework; next, smaller queries fill out images and text. Users can then fine-tune the results or ask the AI to generate more pages via additional prompts.

Tome plans to charge about $10 per user for a monthly subscription when it rolls out its enterprise version this year.

Three decades after its debut, PowerPoint is still a force. Microsoft has invested heavily in OpenAI with a plan to integrate its models across its products. Collaboration software unicorns like Canva, Coda, and Notion are currently announcing AI features on their products.

 

It was fun showing @alexrkonrad what @magicaltome will be able to do in the (not far) future, like generating proposals based on learning your internal material, and offering increasingly insightful advice on how to craft a powerful story.

Check it out:https://t.co/h1PkRm5N5o

— Henri Liriani (@hliriani) February 22, 2023

The most valuable skill in business is about to become obsolete.

Designing a killer slide deck used to take days. Now AI can do it in seconds for anyone.

Introducing Gamma: your AI design partner for presenting ideas beautifully.

Launching today at https://t.co/HrkL4w2vG7 pic.twitter.com/eYpNyn0zpu

— Grant Lee (@thisisgrantlee) March 28, 2023

IBL News | New York

AI startup Anthropic aims to raise $5 billion over the next two years to take on rival OpenAI and enter over a dozen major industries, according to TechCrunch.

Founded in 2021 by former OpenAI researchers as a public benefit corporation, Anthropic says in a pitch for Series C fundraising says that it requires a billion dollars in spending over the next 18 months to build a model with tens of thousands of GPUs, ten times more capable than today’s most powerful AI.

This model — which would be a successor to Claude, Anthropic’s current chatbot — is described as a “next-gen algorithm for AI self-teaching,” making reference to an AI training technique it developed called “constitutional AI,” which aligns with human intentions.

“These models could begin to automate large portions of the economy,” the pitch deck reads.

Other competing startups are Cohere and AI21 Labs in the AI systems space.

Google is also among Anthropic’s investors, having pledged $300 million in Anthropic for a 10% stake.

Other Anthropic backers include James McClave, Facebook and Asana co-founder Dustin Moskovitz, former Google CEO Eric Schmidt, and founding Skype engineer Jaan Tallinn.

IBL News | New York

Silicon Valley–based maker of a dedicated AI computer and the world’s largest computer chip, Cerebras Systems released a series of seven GPT large language models (LLMs), methodology, training weights, and a recipe for open use via the permissive industry-standard Apache 2.0 license. This solution, called Cerebras-GPT, means that these models can be used for research or commercial ventures without royalties.

The company used non-Nvidia GPU-based systems to train LLMs up to 13 billion parameters. All seven models were trained on the sixteen CS-2 systems in the Cerebras Andromeda AI supercomputer using the Chinchilla formula.

“These are the highest accuracy models for a computing budget and are available today open-source,” said the company.

In a first among AI hardware companies, Cerebras researchers trained a series of seven GPT models with 111M, 256M, 590M, 1.3B, 2.7B, 6.7B, and 13B parameters.

“Typically a multi-month undertaking, this work was completed in a few weeks thanks to the incredible speed of the Cerebras CS-2 systems that make up Andromeda, and the ability of Cerebras’ weight streaming architecture to eliminate the pain of distributed computing. These results demonstrate that Cerebras’ systems can train the largest and most complex AI workloads today.”

• “The training weights provide a highly accurate pre-trained model for fine-tuning. By applying a modest amount of custom data, anyone can create powerful, industry-specific applications with minimal work.”

• “The models’ various sizes and their accompanying checkpoints allow AI researchers to create and test new optimizations and workflows that broadly benefit the community.”

Traditional LLM training on GPUs requires a complex amalgam of pipeline, model, and data parallelism techniques. Cerebras’ weight streaming architecture is a data-parallel-only model that requires no code or model modification to scale to arbitrarily large models.

“We’ve worked to make this task easier with releases such as the Pile and the Eval Harness, and we are very excited to see Cerebras build on our work to produce a family of open models that will be useful to researchers around the world,” said Stella Biderman, Executive Director at EleutherAI.

All seven Cerebras-GPT models are available on Hugging Face and Cerebras Model Zoo on GitHub. The Andromeda AI supercomputer used to train these models is available on-demand in this URL.

Cerebras published a technical blog post with the details of the seven models and the scaling laws that they produce. A research paper will be released shortly.

The company posted not just the programs’ source, in Python and TensorFlow format, but also the details of the training regimen by which the programs were brought to a developed state of functionality.

Currently, a handful of companies hold the keys to LLMs. OpenAI is closed, with GTP-4 operating as a black box for the public. Meta’s LLAMA is closed to for-profit organizations, and Google is closed to a varying degree.

Cerebras, echoing the researchers’ community, says that AI needs to be open and reproducible for it to broadly benefit humanity.

🎉 Exciting news! Today we are releasing Cerebras-GPT, a family of 7 GPT models from 111M to 13B parameters trained using the Chinchilla formula. These are the highest accuracy models for a compute budget and are available today open-source! (1/5)

Press: https://t.co/Ltw02PjDQF

— Cerebras (@CerebrasSystems) March 28, 2023

• ZDNet: AI pioneer Cerebras opens up generative AI where OpenAI goes dark

You can now clone ChatGPT!

OpenAI didn’t open-source its models, so we don’t know much behind the scenes.

But the first complete end-to-end model pipeline was just released, and it’s the most practical open-source project resembling ChatGPT.

Here are the details:

— Santiago (@svpino) March 29, 2023

.

IBL News | New York

International Baccalaureate (IB), which offers an alternative qualification to A-levels and Highers, said that their students can use ChatGPT in their essays. But they must be clear when they were quoting its responses.

“Content created by this chatbot must be treated like any other source and attributed when used,” said IB.

IB is taken by thousands of children every year in the UK at more than 120 schools and all over Europe.

ChatGPT has become a sensation since its public release in November 2022, due to its ability to produce plausible responses to text prompts, including requests to write essays.

However, ChatGPT-based cheating capabilities have alarmed teachers and the academic profession.

Matt Glanville, the IB’s Head of Assessment Principles and Practice, said the chatbot should be embraced as “an extraordinary opportunity”.

“The clear line between using ChatGPT and providing original work is exactly the same as using ideas taken from other people or the internet. As with any quote or material adapted from another source, it must be credited in the body of the text and appropriately referenced in the bibliography,” he added.

“When AI can essentially write an essay at the touch of a button, we need our pupils to master different skills, such as understanding if the essay is any good or if it has missed context, has used biased data or if it is lacking in creativity. These will be far more important skills than writing an essay, so the assessment tasks we set will need to reflect this.”

IBL News | New York

OPM (Online Program Management) provider 2U Inc, the owner of edX.org, sued the U.S. Department of Education in federal court this Tuesday over guidance it issued in February governing the relationships between colleges and third-party contractors that perform key services for them.

Hundreds of U.S. colleges use OPM services to start and run online programs, often trading upfront capital from the companies for a portion of their programs’ revenue.

In the filed suit against the Department of Education and its Secretary Miguel Cardona, 2U says that the agency has overreached its authority.

Under the Education Department’s new definition, OPMs that provide colleges with recruiting and retention services, as well as educational content, like 2U, will broadly be considered third-party services.

2U’s lawsuit alleges the department overstepped its power by independently rewriting the Higher Education Act’s definition of a third-party servicer. The suit was filed in U.S. District Court for the District of Columbia.

The agency is particularly focused on entities that receive a share of tuition revenue in exchange for their services, arguing that “it can drive up the price of higher education and draw students to low-value academic programs at subpar institutions.”

“2U cares deeply about our partnerships with leading non-profit colleges and universities across the nation,” Matthew Norden, chief legal officer at 2U, said in a statement. “We believe this recent action by the Department of Education will not only impinge on our ability to serve their students but also ultimately hurt their quality of education.”

According to the lawsuit, 2U would face substantial and irreparable harm if it is classified as a third-party servicer in the eyes of the department, being forced to amend current contracts, undergo “burdensome and intrusive” audits, and pay nonrefundable compliance costs.

The company would also be forced to cut off its South African subsidiary due to the guidance’s ban on foreign-owned and foreign-based subcontractors.