All things AI

[Ditto]

Thread for discussing all things around the rapidly evolving Artificial Intelligence

[Ditto]

I'll start with this link - Google's Gemini article by the BBC.

 

It has to be done - Google are perhaps being a little two-faced with their Gemini representation :mad:

[Ditto]

More seriously, Yr 12 son had online parent's evening yesterday, covering computing and maths. He received a comment/criticism about reusing code. It was his own code and he had built it in a modular way and dealt with that topic admirably!

 

Credit to the teacher for allowing the discussion to proceed, but also focussing on continuing with one's creativity.

 

The teacher asked if we had any questions. I raised the topic of using AI tools for code generation. My son explained many in his class do. The teacher was mildly surprised so many did, but had also explored the topic too. It was a really interesting, if brief discussion, where ultimately the teacher said the skill in using AI effectively is the key. My son said for simple tasks it was OK. The teacher said AI often generated code that wasn't too readable. My son explained for more complex code, going in to some depth with the functional code requirements before the AI is set loose was key. The code generated could then be tweaked - quicker than writing from scratch.

 

I have to give the teacher huge credit for supporting a mature and exploratory approach to coding - there was no 'Don't use AI'. He even admitted to using AI for setting homework for young group lessons. I think they will enjoy the next couple of years!

 

What experiences are others seeing within the school setting - dinosaurs or evolutionary explorers?

[garethEds]

I have been using AI for a while in my department. It provides a great starting point for certain things. I now pay for the 4.5 version of ChatGPT - some of the plugins are great.

 

I was asked to write a policy - so I used GPT to get the initial ideas. This is were I find it works best - getting those initial ideas.

 

Using a PDF GPT I was able to create mock exam paper based on the notes I provide the pupils (on PDF).

 

For a bit of fun, I'm in the middle of writing a python book (In Welsh) using GPT for initial ideas and then editing. Dall-E3 has been used to generate some images. Way from finishing, but the process is interesting. I'll put it on Amazon KDP when finished for 99p.

 

There are a lot of plugins - too many to mention.

 

There is definately a place for it in the running of the school. I found this amusingly interesting: AI headteacher appointed at UK boarding school - CBBC Newsround

 

Gareth

[jthompson]

I don't know how many of my colleagues are using it but I do remember some surprisingly progressive attitudes around it when ChatGPT first emerged a year ago.

 

I don't use it much but may use it for an initial starting off point when doing some coding mini projects.

 

I do wonder about the environmental cost of these generative models, though. Is my using GPT4-Turbo to produce a relatively simple PowerShell script the equivalent of me popping to the shops in an F-22 Raptor?

[NeoNebula]

I stumbled upon your post, and, well, it is quite interesting to read about your experience with ChatGPT and coding. Indeed, using it for the first projects seems legit. On the other hand, I kind of feel your concern about the environmental implications these models bring. Which is a tough balance point for most of us at the crossroads of convenience and sustainability. That might be mitigated by looking into other ways or fine-tuning the use. Some information on sustainable AI design found at: https://appsalon.com.au/artificial-intelligence-design/ if interested.

[jthompson]

I noticed the other day that Google Gemini is now available for Workspace accounts.

[paulkerton]

I noticed the other day that Google Gemini is now available for Workspace accounts.

Yes, the basic version that everyone can access (which will train on your inputs etc) is available to all, but the more advanced models (which won't train on your inputs) are behind the Gemini Add-On.

[Brax]

I’ve been finding Gemini(Still bard to me) is better for writing as you get greater control over the output with drafts and rewriting it to be longer shorter etc… GPT gives more accurate info and is better at coding.

[Bankesy]

Much like OP, we have a Head of IT and indeed school leadership group as a whole who are open to all things AI being used in the right way. We have a small staff IT focus group for all things tech anyway so started discussions in those meetings, I then wrote a policy (using a bit of AI) which got passed down to Heads of Department and now it's just regularly reviewed and updated if for example exam board guidance changes etc.

 

Simply saying no to AI isn't going to get anybody anywhere, in my opinion, staff and students will just miss out on the knowledge they need moving forward. It's about saying yes in the right way, starting small and opening it up gradually to the wider school community.

[6Foot2]

Link: AI to help doctors detect broken bones on X-rays - BBC News

 

Artificial intelligence (AI) has the potential to reduce the number of broken bones missed when doctors analyse X-rays, according to the National Institute for Health and Care Excellence (NICE).

 

The health assessment body says research suggests the technology is safe and could speed up diagnosis, relieving pressure on clinicians and reducing the need for some follow-up appointments.

 

Four AI tools are set to be recommended for use in urgent care in England while more evidence is collected on the benefits of the technology.

 

AI will not be working alone - each image will be reviewed by a healthcare professional.

 

NICE says broken bones are missed in 3-10% of cases - it is the most common diagnostic error in emergency departments.

 

And the trained experts who perform and analyse thousands of X-ray images every day in the NHS are in short supply, with a huge workload.

 

Vacancy rates are 12.5% for radiologists and 15% for radiographers, according to the long-term plan for the NHS in England, external.

 

The solution, according to NICE, is to harness the power of AI to work alongside clinicians.

 

Mark Chapman, director of health technology at NICE, said it would make their job easier.

 

“These AI technologies are safe to use and could spot fractures which humans might miss, given the pressure and demands these professional groups work under," he said.

 

Mr Chapman said AI tools could also potentially speed up diagnosis and reduce the number of follow-up appointments needed because of a fracture missed during an initial assessment.

 

NICE said it was "unlikely" the technology would lead to a rise in incorrect diagnoses or unnecessary referrals to fracture clinics because a radiologist would always review the X-ray images.

 

The process would be better than a clinician looking at images on their own, it said.

 

The potential for using artificial intelligence in healthcare is huge.

 

It is being used already to help detect the earliest signs of breast cancer on scans, spot who is most at risk of heart attacks and predict when the next pandemic will happen.

 

A consultation on the draft , externalNICE guidance, external on this use of AI will end on 5 November.

[Bankesy]

On the subject of AI, our staff recently took a liking to NotebookLM

[ITJAY2023]

Instead of Googling for obscure results like whats that guy in this one movie etc, I use ChatGPT which finds the most insane references you can ever recall.

 

It's also great at sorting lists out, instead of finding a excel formula or creating one I just get it to sort through my csv's.

[Brax]

ChatGPT has gotten quite scary since its “memory” update. As it now uses information I’ve told it in the past to tailor advice I’ve asked it. It is a very useful feature but very scary at the same time.

[Ditto]

The memory feature is very useful. ChatGPT was helping me with a tricky Excel formula afew weeks back, then it became an even more insanely complex problem I needed to solve. It did a fantastic job following up with phase 2. I've also used it recently for 'legal advice' around CGT, legal aid (and how to challenge it's use following unproved allegations) and illegal access to private letters - who can guess I'm supporting someone with a divorce! I actually got to the point where ChatGPT said I used up my allocation and I'd have to start again in a few hours with a new model! The whole interaction felt fully conversational, and in all honesty I felt more comfortable with this than I did speaking to a solicitor, and wasn't costing neigh on £400 per hour. Of course I am aware it's not infallible, but it often referenced other articles, so it left me with plenty of options to cross check. A friend of mine has been a big user of ChatGPT and had signed up to the paid version, using it for work scenarios. Due to a sudden change in circumstances, he has left his latest job and brought his retirement forward a coupe of years. He finds it so useful, he's sticking with the subscription.

[NegativeKillDeath]

I found this interesting regarding the use of AI in job applications Guide for Candidates Applying for Roles at Bedes from Working at Bede's | Job opportunities

 

First time I've seen something like this.

[Rob_D]

Interesting. I can see this becoming more of a thing as the "desperate job seeker" starts cutting corners on their hundreds of applications by using GPT to tailor their covering letter/application, although I wonder how good it would wind up being given it's all going to be rather generic.

 

I like Bedes's approach of guidance, but given the amount of anti-AI sentiment I wonder how long before we start to see straight up bans on using AI for applications? Which would be an interesting one to test in court (can you discriminate on someone based on how they filled in their application? I mean probably yes, but still...).

[jthompson]

Anthropic's Claude can now perform tasks on your computer by viewing what's on the screen and controlling mouse and keyboard input.

 

https://www.anthropic.com/news/3-5-models-and-computer-use

 

"Claude, I'm gonna need you to goo ahead and come in tomorrow to do those TPS cover sheets. Mmmkay".

[Chuckster]

In the past 6-9 months my company has employed a few people in very senior positions, commanding a pay of £85k+ as base salary.

 

Unfortunately, a couple of them have been let go and the rest have had their probation extended.

 

We've come to find that majority have blagged their way through the interview stages (HR + Hiring Manager = useless) and have said they're highly proficient in scripting, which we've found not to be the case. One of the primary scripting abilities we've searched for is the ability to use PowerShell. We're in the midst of automating a lot of things and converting most of the older scripts to PowerShell.

 

Most of the code they've produced has been AI driven, which is fine if you know what's being spewed out.

 

Upon examining these codes, a majority will define variables where it's not used anywhere in the code, or create multiple arrays within arrays that loop many times over when it's not needed. Additionally, it will add some extra lines of code where, for example, it's modifying the registry using Set-Item cmdlet and then equally using another cmdlet that's also doing the same thing.

 

That said, some couldn't explain the code they would produce, which led us to doubt their capabilities.

 

I'm all for AI to lay you a foundation to build upon. I am against it producing the whole shebang which hinders your ability of learning, dissecting and understanding of it.

 

In the summer, we received boat loads of applications from summer interns that you can see have used AI.

 

On one hand, the thought of using AI sounds great, but yet on the other hand, it's sounds like it will make people lazy and give way to those that think doing a crash course in entry-level IT for 6 months will lead to those American type crazy salaries.

[dmj]

In the past 6-9 months my company has employed a few people in very senior positions, commanding a pay of £85k+ as base salary.

 

Unfortunately, a couple of them have been let go and the rest have had their probation extended.

 

We've come to find that majority have blagged their way through the interview stages (HR + Hiring Manager = useless) and have said they're highly proficient in scripting, which we've found not to be the case. One of the primary scripting abilities we've searched for is the ability to use PowerShell. We're in the midst of automating a lot of things and converting most of the older scripts to PowerShell.

 

That's just a bad interview process.

We always get our senior/lead engineers to do the technical interviews. Senior engineers hire the leads in some cases. They can always spot this nonsense.

Management do the informal chats, but they don't run the technical interviews.

[DGardiner]

starting to get bored of all these ai tools teachers are runnig in with now... https://teachmateai.com/

 

what does this add that you couldnt do yourself with the already available and private by default gemini that you all have access too for free....

 

they should be learning to leverage the tools, not paying someone else thousands to write a short bit of context

 

 

BUT MRP SED ITS AMAZING WE NEED IT

[pete]

In the past 6-9 months my company has employed a few people in very senior positions, commanding a pay of £85k+ as base salary.

 

Unfortunately, a couple of them have been let go and the rest have had their probation extended.

 

We've come to find that majority have blagged their way through the interview stages (HR + Hiring Manager = useless) and have said they're highly proficient in scripting, which we've found not to be the case. One of the primary scripting abilities we've searched for is the ability to use PowerShell. We're in the midst of automating a lot of things and converting most of the older scripts to PowerShell.

 

I will happily not lie about my powershell skills in exchange for £85K+/year, provided fully remote is an option.

[dmj]

I will happily not lie about my powershell skills in exchange for £85K+/year, provided fully remote is an option.

 

Why not go for the £120k remote jobs if you're good enough: https://www.itjobswatch.co.uk/jv/GCS-Ltd/Senior-Azure-DevOps-Engineer-Azure-Infrastructure-Architect-Job-Nationwide-UK-3b82vm?jr=8n4-5bn-290

[PotNoodleTech]

Why not go for the £120k remote jobs if you're good enough: https://www.itjobswatch.co.uk/jv/GCS-Ltd/Senior-Azure-DevOps-Engineer-Azure-Infrastructure-Architect-Job-Nationwide-UK-3b82vm?jr=8n4-5bn-290

 

Thanks thats a great job site!

[6Foot2]

Link: Judge Rules in Favor of School That Gave Student a Bad Grade for Using AI

 

A federal judge has ruled that a Massachusetts school was likely within its rights to discipline a student who used generative artificial intelligence to complete an assignment.

 

Jennifer and Dale Harris, the student’s parents, sued the town of Hingham’s school committee earlier this year, arguing that the high school’s policies prohibiting academic dishonesty did not explicitly say that students couldn’t use AI. They asked the federal district court to issue a preliminary injunction that would remove the detention their son received from his academic record and to raise his grade in the A.P. History class where he cheated from a C-plus to a B.

 

“There is nothing in the preliminary factual record to suggest that [Hingham High School] officials were hasty in concluding that [the Harris’s son] had cheated,” Judge Paul Levenson wrote in an order denying the family’s request for a preliminary injunction against the district. “Nor were the consequences Defendants imposed so heavy-handed as to exceed Defendants’ considerable discretion in such matters.”

 

Another student who partnered with the plaintiffs’ son on the project and was also disciplined for cheating did not sue the school district.

 

Levenson wrote that the evidence shows the students didn’t simply use AI to formulate research topics or identify potential sources. “Instead, it seems they indiscriminately copied and pasted text that had been generated by Grammarly.com.” The copied text included citations for non-existent books—which appears to be one of the things AI is best at generating—including a book supposedly written by the author Jane Doe.

 

In their lawsuit, the Harrises said that Hingham school officials pursued a “pervasive, destructive and merciless path of threats, intimidation and coercion to impact and derail [our son’s] future and his exemplary record.” The discipline he received harmed his chances of getting into elite colleges like Stanford University, they alleged.

 

But Levenson said “there is no dispute” that the school officials gave Harrises ample opportunity to express their opinions prior to the lawsuit and that the family hadn’t shown any misconduct by school authorities.

 

Rather than failing to make school policies on AI clear, Levenson wrote, the Hingham High School A.P. English Language teacher gave a lesson on academic integrity and expectations for the use of AI during the first week of a class that the Harris’s son was enrolled in. She also sent students a document that instructed them not to use AI without permission or to copy and paste blocks of text into assignments.

 

While their request for a preliminary injunction was denied and Levenson doesn’t appear particularly convinced by their arguments, the Harris’s lawsuit against the Hingham school committee remains alive.