The churn of experimental AI development, to push innovation and hype attention, perpetually grinds out fresh LLM models that thrill speculators and surprise skeptics. Beyond the hype of DeepSeek, Ollama and Manus etc. our anxieties peek in realisation of it’s insatiable energy appetite.
Media has never been more thoroughly homogenised, distorted and counterfeit at our expense. Shattered context fragments and crumpled shells litter the disinformation pathways. Exploitation of natural resources is on a runaway course in pursuit of unbridled capital gain, whilst a whirlwind of environment catastrophe makes landfall. The daily drama of oligarch drawn maga corporate and transnational capital gorging, depresses our senses, curbing creative expression, suppressing inspirational response, but respond we must!
Fabulous then to visit SDNA prior to the opening of Sentient Currents exhibition of recent image and light sculpture then to hear the delight voiced by all attending for the prompt engineered dish of delicious moths, squid and molluscs @ Bomb Factory 99 Kingsway, London wC2
SPC has limited experience in the art of business development but a substantial record supporting a wide range of socially engaged and exploratory initiatives. The race to re-frame all knowledge, rights and relationships with corporate AI tools has prompted us to call on our friend and local developer Dermot Finn to help mount a response; one that’s more in scale with available resources and in tune with a day to day reality than can yet be offered by corporate business alone.
Our most recent subscriber and contributor EX Networks, operate a contemporary network infrastructure and mission critical hosting at data centres in London and across EU. Their commitment to our AI research initiative RTF, sponsoring some of the time it takes to establish a bank of practical experience to draw from.
Meanwhile, they race within a changing technical environment and accelerating energy costs, so seek innovative opportunities. A growing range of LLM development tools and home brew alternatives to commercial hegemony’s old and new, are in hand.
The word that we were seeking hardware to operate some Ollama AI modelling, got around and as a result, community technology recyclers Catbytes loaned us their best available PC. It has since been topped up with a donated Nvidia 3080 GPU to allow for some modest Open Source LLM model testing via a gigabit fibre host in Peckham. EXN also dropped round an even more chunky PC than we could have anticipated, which features a monster CPU and GPU combo that we placed at a host in Hackney. Jake Bains and Greenman then donated legacy gaming PCs which we reanimated with GPU’s that allow them to run smaller AI models just fine. None match the gruesome capacity nor suck up the energy of a small city like Gemini, OpenAI or Grok megaliths but can all be put into useful service directing agent retrieval and refinement of deeper research prompting and reasoning.
We have since been on a stimulating but sometimes frustrating research journey to brush up on PC hardware configuration and progress with open source LLM deployment with Ollama. The construction, configuration and operation of these fast evolving systems is not such a headache now we gathered some missing info that was not self evident from our standing start in December! Luckily there are many freely available and high quality AI chat systems devoted to problem solving, coding Python for LLM use and organising resources which helped us prepare for more specific application development.
It’s no good relying on the corporate AI systems ‘freely’ available as offered, to continue accommodating the surge in uptake. Better to distribute the compute; establish localised systems to integrate data resources and promote access for social enrichment over commercial abasement.
