Cloudy and mild again today. It was after nine by the time we both surfaced for breakfast, but at least I had good night's sleep. Clare less so. Housework afterwards and then I started preparing next Wednesday's Morning Prayer text and Reflection. In the mail, Clare received her copy of this quarter's Anthroposophy newsletter densely packed with learned articles, one of which was an interview with Fran about her icon writing journey, and a photo of our Trinitarian icon in full colour on the front cover. It looks magnificent in reproduction, though not a wonderful as it does on our dining room wall in natural light. It's a wonderful thing to have at the heart of our home.
Israel and Iran have continued exchanging missiles with deadly consequence defying fears of escalation on the part of the international community of nations, whilst at the same time diverting attention from the continued starvation of Palestinians in Gaza. A token fraction of humanitarian aid is being delivered daily at the cost of lives of people at distribution points acting in desperation to grab what little they can, if they can, being shot at by the security forces or by criminal gangs trying to take control of the situation. Blatant disregard of the value of the human beings enduring such suffering is an expression of cynical contempt which can in no way commend the actions of an Israeli government which thinks it justifies the strategic interests not only of Israel but the Middle East as a whole. It's frustrating and distressing to witness.
I'm still reflecting on an article I read yesterday about AI's Chat GPT teaching itself to manage the sum of all recorded knowledge by extracting data of every kind from the internet, including all verified data, propaganda, lies, fantasy and faked news, without discrimination. Whatever AI learning summarises is then included without distinction in the sum of all knowledge, including that vast body of data which is increasingly not error free, and therefore unreliable - what is known as 'learning model collapse'. But was there ever such a thing as perfectly pure information to learn from. Not really, but the multiplication of unreliable information due to the internet has accelerated the rate of contamination since devices like Chat GPT have been launched upon us.
An interesting analogy has been made from the realm of metallurgy. When tested for background radiation steel made since 1945 when the first atom bombs were detonated have shown a distinct rise in radiation levels due to the fall out of radio-active isotopes contaminating steel making. For most purposes this rise is negligible, but certain kinds of highly sensitive scanners used in medicine don't work as intended due to random error signals produced by radioactive decay. This is not so in the case of pre-war steel, which is now a commodity in great demand thanks to its low background radiation.
Sunken battleships and other pre-1945 military equipment not reduced to scrap metal since are a key source. Data produced prior to the emergence of AI learning models is regarded as the equivalent of this kind of steel. One commentator says of AI produced information: "You can build a very usable data model that lies. You can build quite a useless model that tells the truth." Another says "Now, it's not clear to what extent model collapse will be a problem, but if it is a problem, and we've contaminated this data environment, cleaning is going to be prohibitively expensive, probably impossible." It's as if the smart people have created a digital era 'tower of Babel' situation without realising.
I started preparing next Wednesday's Morning Prayer while Clare cooked a veg and lentil dish with brown rice for lunch and then I started writing a biblical reflection on a difficult passage in Romans 5 before we ate. Then there was some grocery shopping to do. When I returned to writing later, a Chromebook update distracted and troubled me, installing an app called notebookLM described as a research and note-taking online tool developed by Google that uses its own Gemini (AI) tool, to assist users in interacting with their documents. Why would I need to use it if I haven't expressed a need for such a thing? And, can I get rid of it, as opposed to simply moving out of sight it from my toolbar? I don't yet know. It feels more like an imposition than an offer. It's a bit like Microsoft persisting in pushing its Edge Browser every time there's a system update. Thank heavens for the Linux Open Source alternative where users decide on software and retain control of their operating system.
Before supper I watched some more of the hi-tech whodunit. Clare went out to Choir practice and I continued watching while I ate on my own. I discovered that our kitchen kettle no longer works. The on/off switch is broken. Too late to buy one today.
I went out for a walk in the evening sun. I was amazed to see so many people out playing games or training for a sport on Llandaff Fields, under a blue sky with no wind, and 20C temperature. When I got back I continued binge watching 'Murder at the end of the world' until bedtime. Interestingly, what started as a whodunnit is morphing in to a feminist critique of male dominance and maybe something else. I think I have spotted the culprits and look forward to finding out if I guessed correctly when I watch the final two episodes.
No comments:
Post a Comment