Every day, it seems, a new large language model (LLM) is announced with breathless commentary from both its creators and academics on its extraordinary abilities to respond to human prompts. It can fix code! It can write a reference letter! It can summarize an article!
From my perspective as a political and data scientist who is using and teaching about such models, scholars should be wary. The most widely touted LLMs are proprietary and closed: run by companies that do not disclose their underlying model for independent inspection or verification, so researchers and the public dont know on which documents the model has been trained.
The rush to involve such artificial-intelligence (AI) models in research is a problem. Their use threatens hard-won progress on research ethics and the reproducibility of results.
Instead, researchers need to collaborate to develop open-source LLMs that are transparent and not dependent on a corporations favours.
GPT-4 is here: what scientists think
Its true that proprietary models are convenient and can be used out of the box. But it is imperative to invest in open-source LLMs, both by helping to build them and by using them for research. Im optimistic that they will be adopted widely, just as open-source statistical software has been. Proprietary statistical programs were popular initially, but now most of my methodology community uses open-source platforms such as R or Python.
One open-source LLM, BLOOM, was released last July. BLOOM was built by New York City-based AI company Hugging Face and more than 1,000 volunteer researchers, and partially funded by the French government. Other efforts to build open-source LLMs are under way. Such projects are great, but I think we need even more collaboration and pooling of international resources and expertise. Open-source LLMs are generally not as well funded as the big corporate efforts. Also, they need to run to stand still: this field is moving so fast that versions of LLMs are becoming obsolete within weeks or months. The more academics who join these efforts, the better.
Using open-source LLMs is essential for reproducibility. Proprietors of closed LLMs can alter their product or its training data which can change its outputs at any time.
For example, a research group might publish a paper testing whether phrasings suggested by a proprietary LLM can help clinicians to communicate more effectively with patients. If another group tries to replicate that study, who knows whether the models underlying training data will be the same, or even whether the technology will still be supported? GPT-3, released last November by OpenAI in San Francisco, California, has already been supplanted by GPT-4, and presumably supporting the older LLM will soon no longer be the firms main priority.
ChatGPT: five priorities for research
By contrast, with open-source LLMs, researchers can look at the guts of the model to see how it works, customize its code and flag errors. These details include the models tunable parameters and the data on which it was trained. Engagement and policing by the community help to make such models robust in the long term.
The use of proprietary LLMs in scientific studies also has troubling implications for research ethics. The texts used to train these models are unknown: they might include direct messages between users on social-media platforms or content written by children legally unable to consent to sharing their data. Although the people producing the public text might have agreed to a platforms terms of service, this is perhaps not the standard of informed consent that researchers would like to see.
In my view, scientists should move away from using these models in their own work where possible. We should switch to open LLMs and help others to distribute them. Moreover, I think academics, especially those with a large social-media following, shouldnt be pushing others to use proprietary models. If prices were to shoot up, or companies fail, researchers might regret having promoted technologies that leave colleagues trapped in expensive contracts.
Researchers can currently turn to open LLMs produced by private organizations, such as LLaMA, developed by Facebooks parent company Meta in Menlo Park, California. LLaMA was originally released on a case-by-case basis to researchers, but the full model was subsequently leaked online. My colleagues and I are working with Metas open LLM OPT-175B, for instance. Both LLaMA and OPT-175B are free to use. The downside in the long run is that this leaves science relying on corporations benevolence an unstable situation.
There should be academic codes of conduct for working with LLMs, as well as regulation. But these will take time and, in my experience as a political scientist, I expect that such regulations will initially be clumsy and slow to take effect.
In the meantime, massive collaborative projects urgently need support to produce open-source models for research like CERN, the international organization for particle physics, but for LLMs. Governments should increase funding through grants. The field is moving at lightning speed and needs to start coordinating national and international efforts now. The scientific community is best placed to assess the risks of the resulting models, and might need to be cautious about releasing them to the public. But it is clear that the open environment is the right one.
The author declares no competing interests.
Read the original:
Why open-source generative AI models are an ethical way forward ... - Nature.com
- European parliament prepares tough measures over use of AI - Financial Times [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Nvidia stock surges on dominant A.I. market position, buy recommendation from HSBC - Fox Business [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Bloomberg plans to integrate GPT-style A.I. into its terminal - CNBC [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Deepfake porn could be a growing problem amid AI race - The Associated Press [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Workforce ecosystems and AI - Brookings Institution [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Adobe Lightroom AI Feature Tackles a Massive Problem With Photos - CNET [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- How artificial intelligence is matching drugs to patients - BBC [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- These are the tech jobs most threatened by ChatGPT and A.I. - CNBC [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Elon Musk Launches X.AI To Fight ChatGPT Woke AI, Says Twitter Is Breakeven - Forbes [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Two late iconic Israeli singers have been resurrected via AI for a ... - JTA News - Jewish Telegraphic Agency [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- AI anxiety: The workers who fear losing their jobs to artificial ... - BBC [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Grandma exploit tricks Discords AI chatbot into breaking its rules - Polygon [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Commonwealth joins forces with global tech organisations to ... - Commonwealth [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- The power players of retail transformation: IoT, 5G, and AI/ML on Microsoft Cloud - CIO [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- AI is the word as Alphabet and Meta get ready for earnings - MarketWatch [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Purdue launches nation's first Institute of Physical AI (IPAI), recruiting ... - Purdue University [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Will AI ever reach human-level intelligence? We asked 5 experts - The Conversation [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- The next arms race: China leverages AI for edge in future wars - The Japan Times [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Amazon Unleashes Bedrock: The Game-Changing AI Cloud Service Powering the Future of Tech - Yahoo Finance [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Atlassian taps OpenAI to make its collaboration software smarter - CNBC [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Dating an AI? Artificial Intelligence dating app founder predicts the future of AI relationships - Fox News [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Military Tech Execs Tell Congress an AI Pause Is 'Close to Impossible' - Gizmodo [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Philips Future Health Index shows providers plan to invest in AI - Healthcare Finance News [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Reddit Wants to Get Paid for Helping to Teach Big A.I. Systems - The New York Times [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- OpenAIs CEO Says the Age of Giant AI Models Is Already Over - WIRED [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- 9 Resources to Make the Most of Generative AI - WIRED [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Impact of AI on higher education panel event May 3 - Boise State University [Last Updated On: August 18th, 2024] [Originally Added On: April 20th, 2023]
- Microsoft reportedly working on its own AI chips that may rival Nvidia's - The Verge [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Deepfake porn could be a growing problem amid AI race - The Associated Press [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- AI cameras: More than 2 on two-wheelers, even if children, will invite fine - Onmanorama [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- How artificial intelligence is matching drugs to patients - BBC [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- These are the tech jobs most threatened by ChatGPT and A.I. - CNBC [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Will Generative AI Supplant or Supplement Hollywoods Workforce? - Variety [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Marrying Human Interaction and AI with Navid Alipour - Healio [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Competition authorities need to move fast and break up AI - Financial Times [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- 5 AI Projects to Try Right Now - IGN [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Financial Services Will Embrace Generative AI Faster Than You Think - Andreessen Horowitz [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Grandma exploit tricks Discords AI chatbot into breaking its rules - Polygon [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- US FTC leaders will target AI that violates civil rights or is deceptive - Reuters [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Religion against the machine: Pope Francis takes on AI - Euronews [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Fujitsu launches AI platform Fujitsu Kozuchi, streamlining access to ... - Fujitsu [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Commonwealth joins forces with global tech organisations to ... - Commonwealth [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- In this era of AI photography, I no longer believe my eyes - The Guardian [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- AI is the word as Alphabet and Meta get ready for earnings - MarketWatch [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Google CEO Sundar Pichai warns society to brace for impact of A.I. acceleration, says its not for a company to decide' - CNBC [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Purdue launches nation's first Institute of Physical AI (IPAI), recruiting ... - Purdue University [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- We soon wont tell the difference between AI and human music so can pop survive? - The Guardian [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Atlassian brings an AI assistant to Jira and Confluence - TechCrunch [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- How DARPA wants to rethink the fundamentals of AI to include trust - The Register [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Dating an AI? Artificial Intelligence dating app founder predicts the future of AI relationships - Fox News [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- Snapchat expands chatbot powered by ChatGPT to all users, creates AI-generated images - Fox Business [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- ChatGPT sparks AI investment bonanza - DW (English) [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- AI-generated spam may soon be flooding your inbox -- and it will be personalized to be especially persuasive - The Conversation [Last Updated On: April 20th, 2023] [Originally Added On: April 20th, 2023]
- AI predictions for the new year - POLITICO - POLITICO [Last Updated On: August 18th, 2024] [Originally Added On: January 4th, 2024]
- Intel Hires HPE's Justin Hotard To Lead Data Center And AI Group - CRN [Last Updated On: August 18th, 2024] [Originally Added On: January 4th, 2024]
- At Morgan State, seeking AI that is both smart and fair - Baltimore Sun [Last Updated On: August 18th, 2024] [Originally Added On: January 4th, 2024]
- Opinion | A.I. Use by Law Enforcement Must Be Strictly Regulated - The New York Times [Last Updated On: August 18th, 2024] [Originally Added On: January 4th, 2024]
- UBS boosts AI revenue forecast by 40%, calls industry the 'tech theme of the decade' - CNBC [Last Updated On: August 18th, 2024] [Originally Added On: January 4th, 2024]
- AI is here and everywhere: 3 AI researchers look to the challenges ahead in 2024 - The Conversation Indonesia [Last Updated On: August 18th, 2024] [Originally Added On: January 4th, 2024]
- What software developers using ChatGPT can tell us about how it's changing work - Quartz [Last Updated On: August 18th, 2024] [Originally Added On: January 4th, 2024]
- AI and satellite data helped uncover the ocean's 'dark vessels' - Popular Science [Last Updated On: August 18th, 2024] [Originally Added On: January 4th, 2024]
- 2024 health tech budgets to be driven by AI tools, automation - STAT [Last Updated On: August 18th, 2024] [Originally Added On: January 4th, 2024]
- Samsung's new phones replace Google AI with Baidu in China - The Verge [Last Updated On: August 18th, 2024] [Originally Added On: January 28th, 2024]
- Researchers Say the Deepfake Biden Robocall Was Likely Made With Tools From AI Startup ElevenLabs - WIRED [Last Updated On: August 18th, 2024] [Originally Added On: January 28th, 2024]
- Satya Nadella says the explicit Taylor Swift AI fakes are 'alarming and terrible' - The Verge [Last Updated On: August 18th, 2024] [Originally Added On: January 28th, 2024]
- One month with Microsoft's AI vision of the future: Copilot Pro - The Verge [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- Nvidia's Q4 Earnings Blow Past Expectations as Company Benefits From AI Boom - Investopedia [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- HOUSE LAUNCHES BIPARTISAN TASK FORCE ON ARTIFICIAL INTELLIGENCE - Congressman Ted Lieu [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- What is AI governance? - Cointelegraph [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- Scale AI to set the Pentagon's path for testing and evaluating large language models - DefenseScoop [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- Can AI help us forecast extreme weather? - Vox.com [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- Google launches Gemini Business AI, adds $20 to the $6 Workspace bill - Ars Technica [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- AI and You: OpenAI's Sora Previews Text-to-Video Future, First Ivy League AI Degree - CNET [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- Tor Books Criticized for Use of AI-Generated Art in 'Gothikana' Cover Design - Publishers Weekly [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- Generative AI's environmental costs are soaring and mostly secret - Nature.com [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- Energy companies tap AI to detect defects in an aging grid - E&E News by POLITICO [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- Intel Launches World's First Systems Foundry Designed for the AI Era - Investor Relations :: Intel Corporation (INTC) [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- Google Just Released Two Open AI Models That Can Run on Laptops - Singularity Hub [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- AI agents like Rabbit aim to book your vacation and order your Uber - NPR [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]
- The Samsung Galaxy S23 series will get AI features in late March - The Verge [Last Updated On: August 18th, 2024] [Originally Added On: February 22nd, 2024]