The Short, Mid & Long Term Outlook Of AI
I wanted to share how the next phases of AI might play out and explain where we are:
Right Now: We are now in the experiment with, to the semi-automated intersection of AI:
Experiment → Chat - Ask via chat
Task → CoPilot - Action a task (via your copilot like ChatGPT, Claude, Gemini, CoPilot)
Agent Collaboration → Agents - Automate jobs (via agents - like you might want to via team members and act like junior team members)
Colleague Agents → Agent Swarms - Automate sections of roles (agent and bot swarms - talking to you and each other while working through issues)
The Short Term - Trust, Alignment But Protect
More updates, less progress, more switching between ChatGPT, Claude
Meta and Shopify attempting to make Meta Wallet and agentic commerce work
Regulation chat - this won’t go away with the major labs asking for it politically; we will see open source/open weight become more used
Open source models becoming commonplace in normal businesses; smart ones have made the shift
More high-profile hacks - some low-level hacks based on agent use
Trust levels rollercoaster - the more problems that it causes and the more mistakes it makes the more trust declines. If jobs continue to be removed or loss trust will decline. The more complicated or more value it creates, the more trust happens at the top end,
Agent/bot vs person - who owns the problems it causes and how it’s impacting the workplace. There will a lot of people who are fired over what their unapproved agent does
Mid Term - Multi-Agent & Proof Of Right Set Of Products
Multi-agent vs single agent - this use case is still professional but with the likes of Muse, town and Instinct, it’s become more useful for families; except for the multi-agent approach as we do for apps, most will look for one AI provider/agent provider; others will use numerous, and specialists come to the front. “Poly-agents” seems most likely for technical people
Agentic on vs agentic off - so many are pushing for agentic to move forward, so little usage and I can see many blocking; it will cause ops issues
Battle on costs - the race to zero is happening its’s whether normal people will pay or accept costs
Alignment issues - will the LLMs continue to say no to tasks, like Claude refusing to action certain requestions
Hacks - we will see more hacks, more access given and more leaks of information
National security issues - threats, attempts and takedowns will become news items
AGI / Super intelligence battle - the increased noise around hitting, getting to, built AGI/super intelligence
Long Term - Cyber Conflict And Potential Cyber Wars
War and attacks - we have seen reports recently of allowing to much AI involvement, its when there isn’t oversight and human in the loop on defence that there will be huge ripples
Cyber attacks - we are going to see more huge scale cyber attacks and often without any instruction or human driving this. Labs will have to become a fixer and engage in fixing AI issues and oversights
I shared some of these insights on the final episode of AI Moment podcast, it was 15 months of 2x a week 8-10 minute shows. You can listen to the final episode below.
If you’d like to chat through AI, drop me a line to ai@dannydenhard.com
