Now up my dear friend, co-conspirator on @humetricshss.bsky.social, and provost at U of Oregon on the values-enacted peer reviewing process at Public Philosophy Journal #cmuoss25 #openscience

Home - Public Philosophy Journal

The core values of the Public Philosophy Journal, says @cplong.org, are thick collegiality; ethical imagination; diversity, equity, inclusion; and transparency and trust. Find out more publicphilosophyjournal.org #openscience #cmuoss25

Join us for the second half of #CMUOSS25 beginning now! We'll be talking new and evolving forms of evaluation for all kinds of open scholarly outputs with Rachel Kurchin, @cplong.org, and @meredithschmehl.com! #openscience

Where do we draw the line re: #openscience and research being used in ways we don't want via LLMs or elsewhere (like corporate entities monetizing your findings)? #EricYttri takes a more "nihilist" OS is still better for science and if it gets "eaten" by bad actors, it is what it is. #CMUOSS25

Eric Yttri - Biological Sciences - Mellon College of Science - Carnegie Mellon University Eric Yttri

Appreciate #EricYttri guidance to IDENTIFY YOUR NEED/problem first bf developing new #AI tools. Seems obvious but often enthusiasm for the new tech puts cart/horse. #BlockChain anyone?

#CMUOSS25 #OpenScience

Next up #EricYttri long time #OpenScience advocate and CMU faculty - sharing #AI tool developer like his take on the I and R of FAIR principles: "I" can also be integrable and "R" can be "richly described" (strong documentation)

#OpenScience #CMUOSS25

#WhiteHouseAIR&DStrategy - they did a public comment - #OSI partnered with Open Forum of AI which brings academics/researchers into the policy discussion -commented - recommended access to open datasets, development of OS AI - many of @mhmdhsini.bsky.social recs. #CMUOSS25 #OpenScience

#whiteHouseAIActionPlan as relates to #opensource - want full stack AI package run by Dept of Commerce to get ROW to adopt American AI export packages. Doesn't directly mention OS but #OSI will push OS as part of that export package.

#OpenScience #CMUOSS25

Data sharing piece was most controversial in #OSI process for defining #opensource aligning with four freedoms: use, share, modify, study. #OpenData policy of funders like NIH helps inform the approach to pushing for open data in #OpenSource definition for AI: #OpenScience #CMUOSS25

My @humetricshss.bsky.social heart loves @mhmdhsini.bsky.social's call to always return to #openscience values. #CMUOSS25

#OSI wanted a def as a first step to certify if #GenAI is truly "open." V1.0 use, study, modify, share = share source code to train and create data set, complete list of data used, and actual dataset, model weights, and parameters.

#OpenScience #CMUOSS25

Also, not a new problem/question. CC BY on articles has relieved many corporations of subscription costs and allowed them to opt out of supporting the publishing ecosystem.

#openscience #cmuoss25

As #GenAI has exploded, #OSI has looked four freedoms application to AI. What does true OS AI look like: engaged in global co-design process across the globe to determine how to do this, Sloan funded; #OpenScience #CMUOSS25

Up next #OpenSourceInitiative - in 90s focused on approving licenses that meet true OS definition. Use, study, modify, share = "certified open source" - now focus on education policy makers re: what OS is. #OpenScience #CMUOSS25

#OpenScience cannot survive if access depends on corporate servers and private companies. There should be public support independent of corporate structures. It can help us but hinders more the helps given how it's structured right now. Don't forget #OpenScience #CMUOSS25 VALUES.

Much of #GenAI knowledge comes from #OpenScience #CMUOSS25 behaviors - we share openly but corps benefits and don't share but profit. Common knowledge should be transparent and collectively owned. Weakens sense of human agency in knowledge creation process and accountability. #OpenScience

As #genAI footprint increases, independence does NOT come into the ecosystem with the need for specific infrastructure, expertise. As a researcher using truly open models you still rely on infrastructure controlled by a few companies. Corporate dependence undermines spirit of #OpenScience #CMUOSS25

What does #GenAI NOT do for #openscience? OPENNESS - most systems are not actually open -- only open weight but not full #opensource and training data, code, licensing but are closed. You cannot retrain or redistribute.

#OpenScience #CMUOSS25

Many researchers can prompt GPT but few can evaluate for bias, accuracy. These folks are based in rich countries - new #DigitalDivide; Can enhance data, code, protocols but same risks: accuracy, unnecessarily complex solutions, code plagiarism

#OpenScience #CMUOSS25

The second session of our #cmuoss25 open science symposium is underway, with a focus on #AI and #openscience. @cmu.edu

The Use of Open Models in Research | Center for Security and Emerging Technology This report analyzes over 250 scientific publications that use open language models in ways that require access to model weights and derives a taxonomy of use cases that open weights enable. The autho...

#cmuoss25 Mohammad Hosseini #GenAI can accelerate discovery by lowering the bar/cost entry into research -- less $$ needed. BUT it also increases the risk of errors. Moving faster with less patience/time/expertise to verify #OpenScience cset.georgetown.edu/publication/...