From e2a03c96320ed3a69f67a2380e3d4245bca74d71 Mon Sep 17 00:00:00 2001 From: Harvinder Atwal Date: Mon, 25 Mar 2024 11:35:26 +0000 Subject: [PATCH] Added links to images --- .../2024-03-19-aiuk-conf-day1.md | 28 +++++++++---------- .../2024-03-20-aiuk-conf-day2.md | 26 ++++++++--------- 2 files changed, 27 insertions(+), 27 deletions(-) diff --git a/_posts/2024-03-19-aiuk-conf-day1/2024-03-19-aiuk-conf-day1.md b/_posts/2024-03-19-aiuk-conf-day1/2024-03-19-aiuk-conf-day1.md index 006e5e9..f21688a 100644 --- a/_posts/2024-03-19-aiuk-conf-day1/2024-03-19-aiuk-conf-day1.md +++ b/_posts/2024-03-19-aiuk-conf-day1/2024-03-19-aiuk-conf-day1.md @@ -8,14 +8,14 @@ tags: ai Good morning everyone and welcome to AIUK 2024. -![entrance.png](entrance.png) -![westminister.png](westminister.png) +[![entrance.png](entrance.png)](entrance.png) +[![westminister.png](westminister.png)](westminister.png) # A Sneak Peek Inside Isambard-AI `19th March 09:20` First talk is up on one of the most powerful supercomputers built in the UK -![isambard1.png](isambard1.png) +[![isambard1.png](isambard1.png)](isambard1.png) Outline - How did it start? Part of the Uk’s national AI research resource @@ -24,22 +24,22 @@ UX is a key measure of success Sustainable, accessible, It is part of [£300 million to launch first phase of new AI Research Resource – UKRI](https://www.ukri.org/news/300-million-to-launch-first-phase-of-new-ai-research-resource/) -![isambard2.png](isambard2.png) +[![isambard2.png](isambard2.png)](isambard2.png) Building work is still underway. Very optimised for Power usage effectiveness **What is this supercomputer for?** -![isambard3.png](isambard3.png) +[![isambard3.png](isambard3.png)](isambard3.png) Each blade has 900GBs+ of high bandwidth memory 🤯 -![isambard4.png](isambard4.png) +[![isambard4.png](isambard4.png)](isambard4.png) All these layers will be available to use -![isambard5.png](isambard5.png) +[![isambard5.png](isambard5.png)](isambard5.png) It will have guardrails for LLMs to stay within boundaries, eliminate bias and lower toxicity. @@ -58,7 +58,7 @@ Requirements came from a report and it will be iterated on. # Is AGI imminent? `11:05` No. Thanks for coming to my Ted talk. -![agi1.png](agi1.png) +[![agi1.png](agi1.png)](agi1.png) **What is AGI?** Ai is a mirror on ourselves. @@ -109,14 +109,14 @@ Audience questions: # Stalls `13:00` -![stalls.png](stalls.png) -![stalls1.png](stalls1.png) -![stalls2.png](stalls2.png) +[![stalls.png](stalls.png)](stalls.png) +[![stalls1.png](stalls1.png)](stalls1.png) +[![stalls2.png](stalls2.png)](stalls2.png) # AI Culture Long table `13:30` -![longtable.png](longtable.png) +[![longtable.png](longtable.png)](longtable.png) Can AI generate something new? Can AI be good for art? Turning memories into a poem. Opening up access to an experience that wouldn’t normally be available. @@ -134,7 +134,7 @@ Drag performances get a rise in economic downturns # Data, Labour and AI `14:40` Strong start: it seems to be about the abuse of labour and knowledge for AI. -![labour.png](labour.png) +[![labour.png](labour.png)](labour.png) First-hand experience for detoxifying GPT during its training stages. Knowledge was stolen from sites (we knew that). @@ -174,7 +174,7 @@ Unions in the south are connecting with each other. `16:00` Angela McLean, Government Chief Scientific advisor. -![keynote.png](keynote.png) +[![keynote.png](keynote.png)](keynote.png) We need a scientific and systems approach to AI. Systems approach: AI superpower in 2030 diff --git a/_posts/2024-03-20-aiuk-conf-day2/2024-03-20-aiuk-conf-day2.md b/_posts/2024-03-20-aiuk-conf-day2/2024-03-20-aiuk-conf-day2.md index 95d4805..1853428 100644 --- a/_posts/2024-03-20-aiuk-conf-day2/2024-03-20-aiuk-conf-day2.md +++ b/_posts/2024-03-20-aiuk-conf-day2/2024-03-20-aiuk-conf-day2.md @@ -8,14 +8,14 @@ tags: ai Back again for Day 2, right next to the glorious Westminster Abbey -![westminister.png](westminister.png) +[![westminister.png](westminister.png)](westminister.png) # PitchFest `20th March 10:00` -![img.png](img.png) -![img_1.png](img_1.png) -![img_2.png](img_2.png) -![img_3.png](img_3.png) +[![img.png](img.png)](img.png) +[![img_1.png](img_1.png)](img_1.png) +[![img_2.png](img_2.png)](img_2.png) +[![img_3.png](img_3.png)](img_3.png) Researchers (enrichment students) have 90 seconds to pitch their projects and highlighting its real-world impacts, creativity and innovation. They will be judged. We are really trying to get the crowd excited but it’s a tough morning @@ -31,14 +31,14 @@ We are really trying to get the crowd excited but it’s a tough morning # Stalls `11:00` -![img_4.png](img_4.png) -![img_5.png](img_5.png) -![img_6.png](img_6.png) +[![img_4.png](img_4.png)](img_4.png) +[![img_5.png](img_5.png)](img_5.png) +[![img_6.png](img_6.png)](img_6.png) # Autonomous Cyber Defense `14:35` -![img_7.png](img_7.png) +[![img_7.png](img_7.png)](img_7.png) What happens when cyberattacks exceed the capacity for humans to respond? Deep Reinforcement Learning is the major tool for cyber security. @@ -50,12 +50,12 @@ Using AI to find vulnerabilities in systems before humans do. [Purple Llama](https://llama.meta.com/purple-llama/) [AIxCC](https://aicyberchallenge.com/) -![img_8.png](img_8.png) +[![img_8.png](img_8.png)](img_8.png) How do humans oversee a system that runs faster than them? How can we stop them being used against us? Lots of questions around policy issues -![img_9.png](img_9.png) +[![img_9.png](img_9.png)](img_9.png) Questions: **How do you counteract bad actors?** @@ -68,7 +68,7 @@ Less concerned about reward and more on real world systems as opposed to simulat # AI and the battlefield of the future `15:00` -![img_10.png](img_10.png) +[![img_10.png](img_10.png)](img_10.png) Military is looking at AI and it crosses every aspect. Only 40% of mil are available for deployment at any one time. Lots of mil assets ready to be deployed but just waiting and need lot’s of maintenance. Could be an AI problem @@ -77,7 +77,7 @@ Weapons review process for new weapons on the ethics of use. # Digital Twins for the Environment `15:30` -![img_11.png](img_11.png) +[![img_11.png](img_11.png)](img_11.png) Digital replica of a system for forecasting. Difference between that and a simulator? Digital twin can be used by many users and can be continuously updated. Most people believe that you need AI to create a digital twin. It needs real time infusion of data, be accessible to all types of users, highly complex.