California State University System to Become “AI-Powered”
The Chancellor of the California State University (CSU) system has announced that its 23 universities will become “AI-powered.”
What this ridiculous phrase means is that CSU will be partnering with several technology companies in order to encourage and train students and faculty in the use of their products, versions of which will be made available, cost-free, on a “dedicated AI platform.”
If the message wasn’t clear enough, the plan includes establishing an “AI Workforce Acceleration Board” whose job will be to “identify and advocate for AI skills needed in the workplace” and hence, probably, among your course’s learning outcomes.
The initiative was announced in an email to people affiliated with CSU.
Here’s the whole announcement:
I’m curious to hear whether there was any faculty input on this.
This is so effed on many levels.
This shouldn’t need to be said, but a reader cannot sympathize with your frustration if the only thing a reader has to work with is the expression of your frustration.
Sorry, I wasn’t trying to solicit sympathy but will try to return to this with more thoughts.
The short story is that the issues follow the ethics debate over generative AI and LLMs, and they include student privacy risks, IP infringement, unreliability, environmental costs, unchecked AI plagiarism, and more.
All this is beyond the more fundamental, unresolved issues around pedagogy in the age of AI/LLMs, as well as the alleged “shared governance” that’s supposed to exist. I’m not aware that CSU faculty were consulted about this—incl. me as one of the leading AI ethicists in the CSU system—as it’s the first time any of us had heard about this scheme.
Thank you for elaborating your perspective, Patrick; I appreciate it.
This claim is false as a general claim. You know what it is like to feel frustration. More importantly, you just read the letter (one would hope), so surely you do have more to work with than the mere expression. It’s bizarre that you feel that your contribution here is the more reasonable one.
Weird, I am a reader and I sympathize with his frustration. I feel like I’m doing this in a world that is possible.
I share a sense of unease and despair but I’d be curious to hear more about the levels in question.
When you confront the unknown/untested, you don’t need an “acceleration board,” but rather a DECELERATION board. https://youtu.be/_8wLnh7gVLg?feature=shared
Wow. I am so glad I didn’t take a job at a Cal state university last year.
As a professor in the Cal State system, I’m upset on several levels. First, there was certainly no discussion at the campus level, not even a survey to the effect “anybody think this is a good idea?” Second, many professors I know have been waging a war against ChatGPT: Officially embracing it entirely undermines our educational goals. And third (much less seriously), we are a adopting an inferior tool: A better AI was just announced last week!
I wonder whether it would be more productive for us, collectively, not to scream about how “effed” this all is, but instead to advocate for the importance of discussing the ethics of all this, and getting ourselves included in the conversation and teaching of best practices.
Realistically, employers are going to want students who know how to make good use of cutting edge AI, and students are going to want universities that can help provide these employable skills, so universities are like to adopt a lot of sloganeering like this. Standing staunchly against this is probably a losing battle. But trying to steer this in a slightly-less-bad direction might not be…
If you’re referring to my initial comment, that wasn’t an attempt at productivity but an expression of exasperation.
And if you think we need more people “to advocate for the importance of discussing the ethics of all this, and getting ourselves included in the conversation and teaching of best practices”, then you might not have been following this issue (and therefore might not appreciate the many hidden issues). Tons of philosophers have been raising the alarm and doing this since day 1…yet here we are, and hence the exasperation…
You might also want to check your assumption that “employers are going to want students who know how to make good use of cutting edge AI.” Just the opposite, it seems that employers want either (1) AI instead of humans for cost reasons, or (2) humans who can do their own thinking and writing so to have a unique, authentic voice and original analysis—not another self-described “AI prompt engineer” who are a dime a dozen and can crank out more material that sounds the same.
But mostly, I’m not confident at all that employers know what they want in this age of AI/LLMs, just as teachers don’t know how to handle the tech in their courses…
Speaking of what employers want, here’s an article from today on a leading AI company (Anthropic) that doesn’t want its job applicants to use AI…which speaks volumes: https://arstechnica.com/ai/2025/02/irony-alert-anthropic-says-applicants-shouldnt-use-llms
Many companies have banned the use of AI by their employees because of data/IP security, such as Samsung a few months ago: https://www.straitstimes.com/asia/east-asia/samsung-bans-staff-s-ai-use-after-spotting-chatgpt-data-leak
The majority of Big Pharma’s largest companies also ban it for similar reasons; they live and die by their IP: https://www.fiercepharma.com/marketing/two-thirds-top-20-pharmas-have-banned-chatgpt-and-many-life-sci-call-ai-overrated-survey
And here’s a school that doesn’t want human teachers at all but only AI: https://gizmodo.com/arizona-schools-curriculum-will-be-taught-by-ai-no-teachers-2000540905
Lots of other examples online. I’m not saying AI/LLMs are useless, but their utility in the workplace is still TBD and varies from industry to industry. So, broad “employers want” claims about AI should be taken skeptically.
Most employers don’t even understand LLMs or their limits and risks, so they can’t possibly know what they want. Or if all they really want is cheap AI labor, is that an employer you’d even want to work for??
Don’t you realize that blogpost comments about news stories are exclusively for communicating detailed plans of action?!
Funny how you seem to think that employers wanting to shift as much work to AI as they can would contradict them wanting to hire humans who can use cutting edge AI, whereas I would say that it *entails* that they’d want the (relatively few) humans they hire to be ones who can work productively with AI.
So, you’d hinted at a big part of that problem, which is: even if employers want AI over humans, it’ll still need a few humans to babysit the AI, edit outputs, etc. for the foreseeable future. These human jobs will be far fewer than they currently or recently needed. So, graduates will be competing for fewer and fewer jobs.
And what kind of jobs are those anyway? Think of the qualitative differences of being a graphic artist vs. someone who can get AI to crank out art; is the latter really as good or satisfying as the former?
“AI prompt engineers” are already a dime a dozen and will become commodities, unless they have something special to offer. What that special thing is cannot be based on or taught by AI, or else it’s not special when everyone else can also access it. Rather, students who refrain from AI cheating but put in the hard work of learning and writing will have that competitive advantage and differentiation.
Anyway, the goal of an all-AI workforce (if workable) still seems very attractive for lots of companies, given labor costs alone, even if that’s still an aspiration right now.
I am an assistant professor at CSUSB, and I did not know this was in the works (though it is possible the faculty senate was consulted). That aside, the decision troubles me for two reasons.
First, while the letter says these services will be offered to ‘all students, faculty, and staff at no cost,’ it is unlikely these tech companies are offering their product to the university system for free. We are in the midst of a budget crisis, and it would be disappointing to find out that the situation has been made worse by fiat of the chancellor’s office.
Second, the policy will likely lead to worse learning outcomes. One likely outcome of this policy initiative is that more students will consistently outsource writing, exegesis, and critical thinking tasks to Chat GPT. And (I think) the student that consistently outsources such thinking will end up worse at thinking critically and carefully than one that doesn’t.
As an Academic Senator at Sonoma State University, on behalf of lecturer faculty, I do not recall this issue having been raised although my local paper today made reference to an Academic Senate in the CSU approving it last year, I presumed that was Statewide level because it was unfamiliar to me and I usually follow policy closely. I know that in August, we had additionally received an email from our Provost implying we did not have any meaningful and agreed-upon discussion, consultation, or subsequent and agreed upon policy at the level of shared governance.
This was before our University Philosophy Department was cut, which also occurred without appropriate consultation with faculty, and so that context may be meaningful. The acronym CTET, below, stands for the Center for Teaching and Excellence. The content of this email raised many questions for me, but it’s difficult to see in it where there was the sort of agreement now being made, or claimed to have been made, especially in paragraph #4 and the 3rd to last paragraph.
That email, from Aug 21, 2024, states:
“For the last 18 months, we have been tracking the development of generative AI tools, like ChatGPT, Gemini, and DALL-E, and the way they are being used in higher education nationally. Generative AI tools can produce new and unique outputs, including human-like text. They assist in tasks such as writing, research, creating computer code, and language translation, among other tasks. As the availability of these tools increases, so does their usage by students in the educational context. While many of these usages may support teaching and learning, there is growing concern that higher education institutions may see a rise in students using these tools to engage in academic misconduct without proper citation or attribution.
At SSU, during the 2023-24 academic year, the Academic Technology & Instructional Spaces Subcommittee (ATISS), working with support from CTET, conducted a review of AI content detectors as part of a larger pilot to evaluate academic integrity tools, recommending:
That an “AI Content Detection” should not be officially enabled on campus without further guidelines developed by shared governance on its appropriate uses and limitations. The pilot study observed instances where human created text was falsely identified as AI generated, which is potentially highly problematic for any students who could be accused under such circumstances.
Shared governance leadership has indicated that developing a set of AI Guidelines (and possibly policy) is a priority for the coming academic year.
Additional support resources exist through CTET, particularly through their AI in Education page, which includes support information including recommended language for sample syllabus statements for faculty on classroom usage of AI, recordings of workshops on how to address concerns around academic integrity with AI, using AI in the research process, and more.
Additionally:
CTET is currently evaluating a tool, Rumi, that may help to alleviate faculty concerns about academic integrity. Please contact CTET to learn more about opportunities to participate in a Fall technical pilot of the tool.
We encourage those faculty who are interested in exploring the academic integration of AI to reach out to CTET. With our partners in IT, we are also exploring bringing Microsoft Copilot under an educational license to SSU that would protect faculty, student, and staff data from commercial uses and future AI model training.
Faculty should be cautious about using “free” AI tools that have not been licensed for campus use. Such tools do not necessarily meet accessibility requirements and may expose confidential FERPA protected data to third-parties with commercial interests.
In late 2024, Canvas expects to integrate Generative AI for faculty usage in significant ways. These uses can support faculty with assignment creation, writing questions for exams, developing rubrics, and there are even promises of assistance with grading. We encourage those interested to please read this recent press release.
Presently, generative AI features exist in several campus tools including Adobe Photoshop, Zoom, and we expect that many more vendors will be releasing similar features and updates in the future.
In Spring 2025, we anticipate running another instance of the AI in Education FLC. In Spring 2024, 12 faculty joined this community to gain expertise in AI literacy, explored topics in the ethical implications of AI, and developed novel curriculum and assignments utilizing generative AI in their classes. If you are interested in learning more about this, we encourage you to review the recording of last spring’s Faculty Showcase to get a sense of how colleagues have explored working with these tools.
As always, if you have any questions, please reach out to CTET ([email protected]) to schedule a conversation about how to address your concerns about student AI use and/or options for how to integrate AI into your classes.
More work is needed as we continue to investigate these topics and as shared governance begins to develop guidelines on the usage of AI in coursework. We continue to stress, however, the importance of faculty establishing their own positions with respect to the usage of generative AI in their classes. Learning more about the strengths and limitations of generative AI can help to empower faculty to make informed decisions in line with their comfort level.
We want to assure you that both the campus, and the system overall are deeply engaged in crafting comprehensive responses to Generative AI. In June, a CSU-wide working group issued a report on generative AI technologies with a number of recommendations including “Provid[ing] opportunities for faculty and students to co-create GenAI learning objectives, competencies, compacts/agreements, and activities.” We strongly concur with this recommendation. Clarifying your own policies and becoming more familiar with the capabilities and limitations of AI can demystify these tools for both yourself and students and set appropriate boundaries for your teaching on the usage of these tools. I would stress that in this spirit, we continue to experiment with new approaches, treat students with civility, and be aware of the potential ramifications of accusing students of academic misconduct with AI.
The CSU GenAI Resources site contains many useful systemwide resources for faculty training and support. If you have questions or, please see CTET’s AI in Education Initiative or reach out for further consultation.”
Sorry, this doesn’t directly connect to the CSU situation but:
“In late 2024, Canvas expects to integrate Generative AI for faculty usage in significant ways. These uses can support faculty with assignment creation, writing questions for exams, developing rubrics, and there are even promises of assistance with grading.”
As far as I can tell, Canvas still has yet to add basic LMS features like calculated columns and they’re planning to shove generative AI into the platform?!
Am I alone in thinking that some of the angst surrounding this announcement might be overblown? The phrasing of the official announcement suggests to me that the CSU system is trying to embrace the “AI is here to stay, let’s try to figure out how to work with it in the most productive way possible” approach. On the premise that AI really is a genie that cannot and will not be put back in its bottle, I suspect that this might actually be the most rational response.
The tone of “we better make the most of a bad situation” of your description is not apparent in the original message to me at all.
Talking about how your plans will outdo any other university in scale and impact implies valorizing the use of AI.
I certainly understand your point. I suppose it’s language like “impactful, responsible, equitable…”, naming issues like climate change and housing, and so on, that led me to detect a less problematic direction.
Then again, perhaps it’s naive to suspect the presence of anything but administrative calculus behind this lofty rhetoric.
Well, that’s not good.
Great to hear that the CSU system is taking seriously the need to help students learn how to use AI. Programs like this are essential for giving students the skills they’ll need to succeed in jobs after college. They also need to be set up very quickly.
This is gonna give the polisci department at CSULA a collective aneurysm.
this may not be for what you think its for.
this can cut admin costs.
existing models can, e.g do a lot of work that deans and assoc deans do. in the humanities college here, those 2 guys together make $450k. their process is almost abacus-based.
Care to wager on this outcome?
well, the system recently merged 2 campuses. when you merge 2 campuses what becomes immediately redundant is admin, not faculty.
i think there is some awareness that ai can save significant sums in admin/staff.
This will all feel strange at first, just like the first few times you saw a toddler staring at a tablet in a restaurant. But almost all of us will either get used to such things, die off, or fade into cultural irrelevance. We live in a Technopoly, and in a Technopoly there is no resisting new technology once it is available. It will simply be deployed. We did not resist or engage in collective moral regulation when “smart” devices began to invade our collective lives, despite the warnings of prescient critics. We did not resist or engage in collective moral regulation when screens were distributed to young children, habituating them to this new form of life. We did not resist when our departments of education funded the distribution of Ed-Tech into K-12 classrooms all over the world. We offered only whimpers of protest when our papers were scraped for AI training. And for the most part we all continued to incorporate new tech into our lives as it arrived for purchase. Even now, my political energy is being poured into an online blog read by people I don’t know. Though I hope I’m wrong, the complicity in all of this may run incredibly deep and few of us may have the stomach for the alternative, which grows more radical and impossible with each passing year.
From my side, the problem is first and foremost with the nature of these AI firms. They are profit-driven and I don’t see, especially in the US, that anyone, barring the market, would control them. They, on the other hand, control a very risky technology (that is getting riskier because more powerful every day). When universities ally themselves wit these firms, I definitely smell a rat. One would need to know the exact details of the agreement to be able to judge exactly just how much these – in this case – public institutions have enslaved themselves to these firms with not obviously good intentions (from a communal point of view). I fear that just as social media played a large part in destroying/weakening democracy, so will AI have a role in even worse consequences esp if controlled by these companies, and I am not sure I would want my institution have anything to do with them. Of course, it is also true that AI as such is an unavoidable part of our life already now. Still, one can ask the question whether this is the way to go about keeping up with new developments.
I share a lot of the concerns expressed here, but what’s the better alternative? (Not a rhetorical question; I’m honestly asking.)
The alternative was to engage with campus stakeholders to think through this plan, given stated commitments to shared governance. We, the faculty, will be here long after university administrators, presidents, and chancellors move on to their next thing.
I’m not saying that AI/LLMs are entirely useless, but their utility depends on the discipline, e.g., philosophy vs. graphics art vs. computer science. For those disciplines where LLMs are more a cheating aid than a learning tool (like ours), we desperately need guardrails or a pedagogical strategy to help steer students in the right direction. Or if LLMs are the paradigm shift that some folks think they are, then we collectively—not just administrators in a vacuum, apart from their “AI Workforce Acceleration Board” that’s made up entirely of industry advisors—need to rethink the purpose of the modern university.
There’s no evidence that any kind of risk assessment was done before announcing this CSU AI scheme, including on student data security and privacy. Given the for-profit AI vendors involved, it’s not a stretch to think they might want to build profiles of students to be monetized or exploited later, which could make a student unhireable or uninsurable depending on what was revealed. And this is just one of many risks associated with LLMs.
So, it’s possible we could’ve landed on a modified plan, e.g., a dedicated course or even training module on AI literacy and risks before students are given greater access to AI plagiarism tools (which the university would be implicitly promoting if it doesn’t also set up guardrails against cheating).
And there could be other alternatives, but we don’t know because apparently no one bothered to think about them.
Is it really the case that ours is a field where LLM’s are more a “cheating aid” than a “learning tool”? Sure there are some assignments that forbid using LLM’s, for which using them will therefore count as “cheating”. But there are also lots of possible assignments that would allow various constructive uses of LLM’s and lots of possible ways in which LLM’s could be used as learning tools, and I can see strong arguments for thinking that we should shift to these sorts of assignments and learning activities, much as we’ve collectively learned to view spell-checkers and calculators as useful tools that we should encourage our students to use, rather than the “cheating aids” that they initially were viewed as.
(I’m totally with you in thinking that faculty should be thoroughly included in this process, not just administrators and industry leaders.)
“But there are also lots of possible assignments that would allow various constructive uses of LLM’s and lots of possible ways in which LLM’s could be used as learning tools.”
Perhaps, but no one has cracked this nut yet. Or if you think you had, feel free to share your assignments or ideas!
You could, sure. That’s just not how students are using it.
Thank you, Patrick!
“For those disciplines where LLMs are more a cheating aid than a learning tool (like ours), we desperately need guardrails or a pedagogical strategy to help steer students in the right direction.”
Thank you, Patrick.
“For those disciplines where LLMs are more a cheating aid than a learning tool (like ours), we desperately need guardrails or a pedagogical strategy to help steer students in the right direction.”
Speaking only for myself, I’ve found these tools to be quite helpful pedagogically. Learning from my LLM has become part of my morning routine. So although I had your response initially, I no longer think of them as more a cheating aid than a learning tool, even in our discipline, and I’m discovering ways to implement them in assignments.
Yes, we need guardrails and a strategy, but I also see in the collective response a lot of indignation and even panic that I think won’t age well and possibly would have made things worse than they are. See r/Professors.
I’m not lumping you in with that, but maybe you’re more sympathetic with them than I am. Was there outcry over free Zoom or even Outlook, which are just as capable of building profiles of students to exploit later? If so, I didn’t see it.
Again, I agree with most of the concerns here, but if “so effed on many levels” is a 9, I’m more of a 6 or 7.
This is my entire message. Not sure how the fragment above got posted, and I can’t delete it.
I agree that it seems possible LLMs can be used productively in the classroom; but as I said before, no one apparently has cracked that nut yet.
If you think you have, I’m sure we’d all be curious how. That is, could you elaborate on this bit?:
“Learning from my LLM has become part of my morning routine.”
Not sure exactly what would count for you, but I’ll share what I’ve done that I found productive.
In fall, we did a kind of Turing test to test the claim some people still make that they can detect AI writing. (They can’t.)
I had (verbal) conversations with my GPT in class (an AI seminar) and let my students converse with it, too. Students asked it about machine consciousness, asked it to justify its existence, asked it why we should trust it and the tech companies behind it, asked about cheating and AI reliance, and many more interesting questions. We later discussed its capacities and limitations. In the following class we listened to NotebookLM’s podcast about our previous class (based on our Zoom transcript). I didn’t tell them this was AI, and no one who didn’t already know about NLM had any idea whatsoever. While this last use was partly for fun, all these uses where very helpful in dispelling illusions people have about AI.
One optional assignment was to write a research presentation with the help of an LLM and to turn in all the dialogue with the research project and along with an evaluation of the LLM’s contributions or lack thereof.
Lately, I’ve been recording philosophical conversations I’ve had with mine about, say, how Strawson would respond to Nahmias’s views on free will and bypassing. I now plan an assignment where students must record a productive philosophical conversation (under a number of practical constraints) with an LLM on a selected issue and analyze it later. Students will be evaluated on how they craft questions and evaluate the LLM’s replies.
I highly recommend the wonderful post by Sophie Nelson in the thread about how we use these things. I have been doing very much the same thing, and I hope to model this kind approach for my students.
I will add that students and faculty who do not speak English as their first language have raved about how helpful it has been since it removes certain psychological barriers to asking for help.
Again, I have many concerns, and we have an obligation to distinguish original work from merelow/zero-effort GenAI products.
But we also have an obligation to prepare our students to thrive and live responsibly in a world of advancing AI. Making these resources available to students is in itself not the bad part.
Thanks, James, but that seems less like pedagogy for a philosophy classroom than slipping in some AI literacy into your class. Specifically, the big trick everyone seems to be struggling with is with student assessment—the necessary evil of our profession.
Seems that all the things you make your students submit could also be AI generated, which just kicks the can down the road, as it were. Sure, we might have an obligation to distinguish original work from AI work, but as you pointed out, that is increasingly difficult to do. And it’s a serious burden on overworked teachers already. So, part of good pedagogy here is sustainability for teachers, i.e., to not have to kill ourselves (more) in running a class.
I did see Sophie Nelson’s interesting comments on the other DN thread. Those seems to be decent ideas for self-improvement, but again I don’t see how those translate into assessable student activities. And if you encourage students to use AI to that great extent, then there’s not much stopping them from taking the extra step of using AI to generate entire drafts (or final versions) of written work.
But I do appreciate those ideas, which seem to be getting more creative and useful as the tech develops. I can see how a dedicated scholar (grad student, faculty, etc.) could use these tools responsibly, which depends on understanding LLMs limits and not over-trusting them; but that’s not a realistic expectation for undergrad students, esp. non-majors who don’t care as much.
Finally, I might agree with your last bit that “Making these resources available to students is in itself not the bad part.”
But in the existing context, it is bad. It’s bad because it’s promoting the use of AI without also offering solutions or guardrails to curb AI cheating. It’s bad for its massive environmental impact. It’s bad for how it was rolled out without “shared governance” or any real faculty input. And so on.
I see what you were after now. I think OpenAI and other platforms are morally obligated to archive everything they produce so that (under morally permissible conditions) whether something is AI-generated can be confirmed TurnItIn-style. For all I know this or something like it is or will be part of GPT Edu. I have little doubt that most platforms already keep most exchanges. This would only mitigate, of course, but done right it could be significant.
In the alternative scenario, CSUs don’t make these tools available, but (genie’s out) students use them anyway and we have no way of tracking that use. Some students can afford the $19/mo or even $200/mo, while others rely on the free versions. Less potential for oversight, more inequalities and unfairness.
The CSU has been a colossal joke for a long time, so I’m not at all surprised to see that they’ve finally decided to make it official.
Here’s a statement this afternoon from the CSU faculty union (CA Faculty Assoc.). It also focuses on faculty issues, such as IP, surveillance, academic freedom, as well as shared governance.
——————————————
Dear Colleagues,
You’ve likely received a message from the Chancellor’s Office or your campus about their new artificial intelligence (A.I.) initiatives. This announcement proclaims significant investments in the use and expansion of A.I. technology on campus and in our classrooms. While we recognize there are multiple perspectives on the use of A.I, we have significant concern about faculty jobs, academic freedom, intellectual property, and surveillance. Any integration and use of A.I. in the classroom must be led by faculty, not by administrators and tech companies. Faculty should have the power to decide how and whether to use these tools and should not be subject to repercussions for using A.I. in responsible ways, nor for refusing to use it.
We also know the many issues with A.I. beyond the impact on our work. There is the profound environmental toll of A.I., from significant electricity use to increasing water consumption to cool servers. Amazon workers have complained that A.I. bots fired them without human interaction. Surveillance apps track how often workers leave their desks. A.I. tools for resume screenings have been notorious for discriminating against people based on race and gender. A.I. models have repeatedly been used to create racist imagery, perpetuate racial discrimination in policing, and the A.I models themselves are engaging in racist behavior. Companies are also stealing people’s original work. Their art, literature, essays, music, data, etc., are being used to train A.I. to reproduce and replace it. It is being used to replace mental health counselors. The list goes on.
Management decision making is directly responsible for cuts, loss of work, and layoffs. At the same time, they continue to invest university resources to launch initiatives with little to no input from faculty, students, and staff.
We will keep you informed as we demand to meet with CSU management to discuss how this new initiative impacts faculty work. Faculty intellectual property is protected by our contract. This includes any and all work produced as we perform our duties, from video recordings and tutorials to lecture notes, from research and creative activities to syllabi, from classroom assignments to athletic regimens. Unless faculty sign a separate agreement detailing how management can use faculty IP, neither the CSU nor third parties have a right to the work we create.
Make sure to sign up for Headlines to stay up to date on CFA news. If you are not a CFA member, we encourage you to join. We can only maintain a quality higher education by working together.
In solidarity,
CFA Officers
I guess this will be the natural progression of AI in education and is ok as far as content is double checked because sometimes it’s wrong info (I had a couple experiences) . Also in the issues to tackle they mention climate change, and most AI uses lots of energy , water and chemicals to cool down