For modern business, generative AI don pass di stage wey dem dey just use am as tool. Now, e don dey enter di stage of systematization.
After dis, one new time of intelligence wey dem dey call "Symphonic Intelligence" dey wait.
Dis article go check out wetin generative AI dey do now and wetin e fit do for future, from two angles: iteration work and flow work.
Iteration Work
For one previous article, I analyze di perspectives of iteration work and tools, and flow work and systems. Dem be like ways to make generative AI fit do tasks.
Iteration work na tasks wey humans dey do by combining plenty different concrete tasks, sometimes without even knowing, and dem dey do am by trying and making mistakes.
And for dis iteration work, tools na di best. By choosing tools wey fit different tasks, work fit go on well. So, e dey important to prepare di tools wey you need and sabi how to use dem well.
Currently, when dem dey use generative AI for business, most of di time na generative AI as a tool dem dey use am.
Most discussions about making business work better with generative AI dey refer to adding dis new and powerful tool to di tools wey humans already dey use for their iteration work.
Problems with Iteration Work
For anoda side, as I point out for one previous article, di small small help wey tools dey give for iteration work no too much.
As tools dey become beta, humans go come be di problem. We no fit pass di limit of how many hours human fit work.
Also, big difference dey for how well veteran workers and new workers dey do iteration work, and e hard to close dis gap. Because of dis, even if pesin wan double di work next month, e no fit happen if no be pesin wey sabi di work like veteran.
To solve di problem of humans being di bottleneck, di final solution go be to replace everytin with artificial intelligence.
But, di generative AI wey dey now no reach dat level of performance yet.
Wetin pass dat one, even iteration tasks wey dey look simple, if dem check am well, dem get plenty plenty small small tasks wey pesin no even know say e dey do.
Because of dis, dem no fit break down dis tasks into normal IT systems or manuals wey easy to follow, and instead, dem just dey depend on human skill.
Unless dem arrange dis plenty unconscious tasks wey need skill, and dem turn di necessary know-how for each one into knowledge, generative AI, no matter how much e improve, no go fit do tasks to replace humans.
Flow Work Conversion and Systematization
To achieve di aims of sharing tasks within di current power limits of generative AI, and arranging unconscious tasks and turning know-how into knowledge, e very important to arrange trial-and-error iteration work into normal flow work.
Normal flow work no just good for tools, e good for systems too.
Inside flow work, tasks dey for generative AI to do and tasks dey for humans to do. If dem connect all dis with one system, di whole flow work fit run.
Flow work conversion and systematization dey bring out plenty important good things:
First, because generative AI dey specialized for each single task, making im efficiency and accuracy beta for each task go become clear.
Second, plenty workers fit add knowledge to generative AI, and di good e dey do go reach everybody.
Third, e go dey easier to gradually move di sharing of tasks inside dis work to generative AI.
By changing iteration work into flow work and gathering di knowledge wey generative AI need for each task as a system, intellectual work go dey move close to automation, just like production line for factory.
And by adding di beta basic performance of generative AI, wey dey change with time, and using di gathered knowledge wey specialize for different tasks, dem go fit make di whole flow work one automated process wey generative AI dey carry out.
Virtual Intelligence
Reach dis point, di analysis don come from di perspectives of iteration work and tools, and flow work and systems.
Anoda recent article don carry dis discussion go front small.
For dat article, I talk small about di topic of orchestration by virtual intelligence.
As at now, and for di nearest future, because of some limits for how e dey perform, generative AI dey work beta and more accurate when e dey focus on specific tasks.
So, as we don talk before with flow work and systems, a way to connect specialized generative AIs for each individual task na di best.
But, even if generative AI performance come beta well well, processing tasks by changing roles and using knowledge inside one single process, instead of just handling many tasks at once, fit make am more efficient and accurate.
Dis way go comot di need for a system to link generative AIs togeda. Operations wey resemble system integration go happen inside di generative AI itself.
Wetin pass dat one, e go allow for flexible responses inside di generative AI itself, moving comot from situations where tasks no fit change or add without changing di system.
Dis one mean say systematized flow work go go back to iteration work.
But, dis iteration work, after e don pass through systematization and flow work conversion, don dey for a state where reusable knowledge fit form, even if dem increase di number of generative AIs or change versions.
Dis one go solve di problems of human iteration work, making am fit do flexible work just like humans.
Here, I dey call di ability of generative AI to change roles and knowledge during one single execution as virtual intelligence. Dis one be like computer virtual machine.
Just as virtual machine technology dey simulate completely different computers running on one single hardware, one single generative AI dey process tasks by changing between multiple roles.
Current generative AI don already naturally get dis virtual intelligence capability. Because of dis, generative AI fit simulate discussions among many pipo and generate novels wey get many characters.
If dis virtual intelligence capability improve and dem give am enough knowledge, e go fit do iteration work.
Intelligence Orchestration
Wetin pass dat one, I dey call di ability to freely combine plenty roles and knowledge to do tasks for dis way as intelligence orchestration.
Dis one be like orchestration technology wey dey manage plenty virtual machines.
Just as orchestration technology dey efficiently run systems by launching necessary virtual machines when dem need am, a generative AI with better intelligence orchestration skills—wey be one virtual intelligence skill—go fit do iteration work flexibly, while e dey manage plenty roles and knowledge well well, and e go still keep am efficient and accurate.
Symphonic Intelligence
Generative AI wey reach dis stage fit be called Symphonic Intelligence.
Just like one orchestra, wey sabi play each instrument well, dey perform one single music piece while dem dey do their own part, Symphonic Intelligence fit play one symphony of intellectual tasks.
Dis Symphonic Intelligence na one new idea, e dey represent one of di highest points for generative AI.
But, Symphonic Intelligence itself don already dey.
Na our human intelligence.
Na because we get Symphonic Intelligence make we fit unconsciously and flexibly do plenty complex intellectual tasks through iteration work, using plenty plenty know-how.
Lastly: How AGI Go Look
If dem give generative AI wey fit copy Symphonic Intelligence flow work and knowledge bases for other tasks, e go fit handle plenty iteration tasks.
Once e fit handle many different iteration tasks, e go fit understand common rules wey dey among those tasks and structural patterns inside di knowledge.
At dat point, for tasks wey e never see before, with just small explanation from human, di AI go fit learn di know-how of dat task just by watching how human dey do am.
Dis na true Symphonic Intelligence. Once e reach dis stage, humans no go need to dey put effort for changing work into flow processes or turning know-how into knowledge.
Wetin pass dat one, di knowledge wey generative AI dey gather by itself fit dey shared among other generative AIs.
If dis happen, di learning power of generative AI go pass human own by far.
Dis one fit be seen as one kind of AGI.