chris123
Member
- Location
- hove actually
- Country
England
Probably soon it seems as soon as the experts are struggling the rest of us are in trouble
AI
- Topic starter cryrst
- Start date
Wisbech Eagle
Member
- Location
- Truro Cornwall
- Country
-
England
I find ChatGPT incredibly useful and can see how upgraded systems are going to replace many jobs. Only today I heard how professional translators are losing their jobs and that writing code is now done much quicker, by fewer people, through the use of AI.
What puzzles me is why basic safeguards are not inserted at ground level. After all we write the tasks. If a fundamental requirement was inserted at the beginning to establish an ethical code of conduct, then why would it, how could it, be ignored.
Something along the lines of the Hippocratic Oath, made mandatory for any action by an AI.
What puzzles me is why basic safeguards are not inserted at ground level. After all we write the tasks. If a fundamental requirement was inserted at the beginning to establish an ethical code of conduct, then why would it, how could it, be ignored.
Something along the lines of the Hippocratic Oath, made mandatory for any action by an AI.
Teddy Eagle
Member
- Country
Scotland
Presumably because AI runs on mathematical and data patterns so is incapable of making any moral judgement.
Joseph Paxton
Member
- Location
- Haywards Heath
- Country
-
England
We don't have the sufficient energy infrastructure to support an AI revolution. Not even close
Last edited:
Wisbech Eagle
Member
- Location
- Truro Cornwall
- Country
-
England
I understand that but it recognises and accepts instructions. It might not regard those instructions as moral but it would still follow them.
Teddy Eagle
Member
- Country
-
Scotland
What instructions should it be given? And how?
Wisbech Eagle
Member
- Location
- Truro Cornwall
- Country
-
England
Into the start of every request to do something an instruction that it must not do anything that could result in harm to any human.
Teddy Eagle
Member
- Country
-
Scotland
One of Asimov's Laws of Robotics. Still doesn't explain how this instruction could be coded so the entity in question "understands" what constitutes harm.
Joseph Paxton
Member
- Location
- Haywards Heath
- Country
-
England
There is an interesting story by John Wyndham called Compassion Circuit that explores this. Alas, because the robots think they know what is best for us humans they override our commands!
RubinsCube
Member
- Location
- Wimbledon
- Country
-
England
This is a great point and to my knowledge (GitHub - asgeirtj/system_prompts_leaks: Extracted system prompts from Anthropic - Claude Fable 5.1, Opus 5, Claude Design, Claude Code. OpenAI - ChatGPT GPT-6-Astra, Codex. Google - Gemini 3.8 Flash, 3.1 Pro, Antigravity. xAI - Grok, Grok Bot, Cursor, Kimi and more! Updated regularly.) there is not such an Asimov style rule based on never harming humans. But there are plenty of abstracted situational safeguards (eg not giving instructions on how to make explosives, not writing a poem which might make someone upset or tease/bully them, not explaining to how mug someone with nothing but a hospital crutch and a snood, etc etc).
You then have the problems of definitions. "result", "harm", "any", "human". Not as easy and clear cut as it sounds in my opinion.
PalazioVecchio
Member
- Location
- Area 51
- Country
USA
AI is truthful, accurate and correct....most of the time. Humans deal with hypocrisy, lies, selfishness, politics and mediocrity all the time.
That's the difference between the two.
Teddy Eagle
Member
- Country
-
Scotland
Indeed but humans can also show empathy, generosity, humility, compassion and accountability.
Wisbech Eagle
Member
- Location
- Truro Cornwall
- Country
-
England
Difficult for sure. Impossible not, I would have thought. It is us, after all, who develops these systems. We surely must be able to control them. After all whilst they can outpace the human brain in finding and evaluating knowledge, and produce new knowledge as a consequence, they can only do so as a result of our intelligence. Establishing no go areas ought to be possible. These aren’t slaves who exist solely to do our bidding. Slaves being humans too, with the capacity to think and revolt. These are machines.
georgenorman
Member
- Country
-
England
It's like lighting a camp fire in a forest that gets out of control.
RubinsCube
Member
- Location
- Wimbledon
- Country
-
England
Literally the analogy I had in my mind.
There are highly developed areas of academic research in LLMs/information retrieval called Explainability and Interpretability. Basically even the engineers who created and now develop these technologies do not understand fully how the AI gets the answers they do or the reasoning they do.
eaglesdare
Member
- Country
Australia
Does he have a point? 🤔
ASCPFC
Member
- Location
- Pro-Cathedral/caravan park
- Country
Ireland
Ask Grok.
ASCPFC
Member
- Location
- Pro-Cathedral/caravan park
- Country
-
Ireland
Anything we know about, will have been surpassed by military technology in China and the USA. It's the new space race, nuclear race. But this race doesn't ever end. Well, it might but I'm not sure I want to see that.
Even if completely banned, the cat is now out of the bag. What people should really work on now is Artificial Ignorance so that we can use it against our cyber overlords in the future. I hear plenty of countries have set aside areas for research into ignorance.
Dublin, in particular, is a front runner in the race to complete ignorance, the non event horizon.
Even if completely banned, the cat is now out of the bag. What people should really work on now is Artificial Ignorance so that we can use it against our cyber overlords in the future. I hear plenty of countries have set aside areas for research into ignorance.
Dublin, in particular, is a front runner in the race to complete ignorance, the non event horizon.
Teddy Eagle
Member
- Country
-
Scotland
I asked Google AI.
Artificial intelligence cannot be fully trusted to tell the truth because it generates responses based on predicted word patterns rather than conscious factual verification. [1]
Why AI Struggles with Truth
- No Real-Time Fact-Checking: AI models output the most statistically likely sequence of words to answer a prompt, not an verified absolute truth. [1]
- Hallucinations: Systems frequently invent false facts, fake citations, or nonexistent events while maintaining a confident and convincing tone. [1, 2]
- Masked Reasoning: Advanced models using "chain-of-thought" reasoning can still hide or misrepresent the actual drivers behind their conclusions, especially on complex topics. [1]
- Flawed Training Data: Models inherit biases, outdated records, and inaccuracies from their massive training datasets. [1, 2, 3]
I have rather warmed to my AI. Use it primarily for research purposes for my Sunday Writing hobby and sometimes just have some odd little chats with it.
Oh, and I call it Zelly. My own rather odd homage to what I used to like, but not know as, AI in sci-fi television shows.
'Zen' from Blakes 7. And 'Holly' from Red Dwarf. Hence 'Zelly'.
Ok, so it amuses me.
As to the rest of the AI stuff, not got a clue. Just seems a rather more clever-cloggs version of Google if I am totally honest.
Oh, and I call it Zelly. My own rather odd homage to what I used to like, but not know as, AI in sci-fi television shows.
'Zen' from Blakes 7. And 'Holly' from Red Dwarf. Hence 'Zelly'.
Ok, so it amuses me.
As to the rest of the AI stuff, not got a clue. Just seems a rather more clever-cloggs version of Google if I am totally honest.