Guide­lines, more than policies

Gen­er­a­tive AI has been on every head­line for the past cou­ple years. We’ve toyed with many tools and expe­ri­enced, in one way or anoth­er, that feel­ing of wit­ness­ing mag­ic. At the same time, there are sev­er­al con­cerns about the use of AI in our indus­try, rang­ing from copy­right infringe­ment and pla­gia­rism, to sim­u­la­tion and data man­age­ment, among many others.

Ethics is ulti­mate­ly, about bound­aries, the lines we draw to make sure we oper­ate with­in the lim­its of what we deem accept­able. And we work in an indus­try where peo­ple put their mon­ey where our mouths are, so what we say mat­ters, and whether or not we can be trust­ed mat­ters too. We deal with trust, and that is main­ly built with honesty.

The tool or the machine

These times are push­ing us to face the age-old chal­lenge of mak­ing our­selves irre­place­able. We do this by know­ing more about our craft, not less, by under­stand­ing the real changes in our work, and by valu­ing tal­ent and exper­tise. Our craft is a human thing. What we do is about peo­ple. Our job will always be the human part of the job. 

We use tools, that we cre­ate, to cre­ate fur­ther, and to make things beau­ti­ful and use­ful. But those things reflect us and how we behave, what we do. And what we do is exper­i­ment, learn, adapt, go back to what oth­ers did before, and draw from expe­ri­ence. Connect.

In your face, Skynet.

A pro­posed approach:

• Set rules.
• Be open about them.
• Ques­tion them.
• Iter­ate.

While we fig­ure out what these changes mean and wait for those in pow­er to catch up with this cen­tu­ry’s zeit­geist, we can bear our respon­si­bil­i­ty, define our cri­te­ria, set our lim­its and make them crys­tal clear for cus­tomers, part­ners, and providers.

This ​“AI Code of Con­duct” can take many forms. Here’s what we’re doing at One Totem:

Ask­ing ques­tions to fig­ure out the ​“why” behind what we are doing.

When using these tools, we ask ourselves:

• Are we automat­ing tedious tasks? Or replac­ing someone?
• Is the result bet­ter than doing it the good old-fash­ioned way?
• Who are we help­ing? Clients, staff, our­selves, the world?
• Do we have the resources (time, bud­get) to keep peo­ple in charge of deliv­er­ing creativity?
• Are we opti­miz­ing some­thing that does­n’t need it?.

Tech is only a means and nev­er an end.

If it’s to achieve more with less time and effort, we must think care­ful­ly about who that ben­e­fits, and be trans­par­ent about it. The race to obses­sive effi­cien­cy does not gen­er­al­ly ben­e­fit cre­ativ­i­ty, explo­ration or innovation.

If it will impact oth­er peo­ple’s lives neg­a­tive­ly, we’ll con­sid­er if there’s any oth­er way.

If it’s because ​“Every­body’s doing it so why can’t we” we say that fol­low­ing trends is a per­son­al choice, and there’s no real val­ue in being an ear­ly adopter of anything.

If it’s because the result­ing work real­ly improves all the oth­er very human stuff we do, we make it work (and be trans­par­ent about it: we did­n’t cook the chick­en, we ordered delivery).

Being trans­par­ent about what we are doing.

We tell our cus­tomers when some­thing was built with AI. Arti­cles, images, videos, descrip­tions, cus­tomer sup­port, order pro­cess­ing, live chat, prod­uct descrip­tions, search sug­ges­tions, etc. We’re open about the tools we use, and about why, and how we use them.

What we use it for

Dou­ble-check­ing stuff

Cre­ative writ­ing remains exclu­sive­ly in the realm of our team’s imper­fect and won­der­ful organ­ic brains.

How­ev­er, most of our team mem­bers work in a for­eign lan­guage on a dai­ly basis. Some of us in more than one! For gen­er­al com­mu­ni­ca­tion and oper­a­tions, we use AI tools for basic gram­mar, spelling and style check, and plan­ning lay­out or struc­ture. We still use Hem­ing­way and Gram­marly for this, too.

Automat­ing non-val­ue tedious tasks

We still write our emails. We use AI tools to sum­ma­rize con­ver­sa­tions in order to pre­vent mis­com­mu­ni­ca­tions. We are using it to sug­gest Excel for­mu­las, write legal dis­claimers and unortho­dox lorem ipsums or do basic gen­er­al research. And for fun too. We use it to choose ran­dom words for our week­ly remote Pic­tionary team games.

Cod­ing

Though we don’t use it as much as we could, it has proven use­ful to find bugs and sug­gest workarounds or solu­tions. We’re still scratch­ing the sur­face on this. Truth is, it’s becom­ing a sub­sti­tute for Stack­Ex­change. Archi­tec­tur­ing or design­ing solu­tions will stay with Team Human.

Illus­trat­ing

This one treads the fine line of the eth­i­cal dilem­mas we dis­cussed ear­li­er. We like to exper­i­ment, and we don’t have the bud­get, time or tal­ent to paint cryp­tic 60s sci-fi cus­tom images for our blog. The results are usu­al­ly far from what we ini­tial­ly envi­sion, but still cool and dis­tinc­tive enough. We use AI gen­er­at­ed images to illus­trate our blog post and newsletters.

Debat­ing

“Hey, Hal” (we like nam­ing it), ​“I’m con­fi­dent I can prove that Rock Lob­ster is the best song ever writ­ten. Please coun­ter­ar­gue!”. It’s use­ful for find­ing gaps in your own nar­ra­tive, or ques­tion­ing pre­con­cep­tions. This is par­tic­u­lar­ly use­ful at 2am when you can’t find any­one ready to con­front you or bounce ideas off.

Bridg­ing ideas

You know when you have this intu­ition that there are cer­tain pat­terns between seem­ing­ly sep­a­rate things, but you don’t know enough to find them? Future Ultron does. It can join the dots faster. Find a few edges and ask the bot to find the com­mon ground, and jus­ti­fy the argument.

Run­ning stats

We’re look­ing into using Auto­G­PT to run data and ana­lyt­ics check-ins on our prop­er­ties, build­ing cus­tom reports with tai­lored data peri­od com­par­isons and auto­mat­ed for­mat­ting. Still in the works, we’ll share when we get there.

We don’t (and won’t) use it for:

Fak­ing insight

Any and every opin­ion we express is ours, backed up by our per­son­al and group ideas, expe­ri­ence, knowl­edge and crit­i­cal thinking.

We do not use it for brain­storm­ing. We don’t believe in using soft­ware to find us ideas. Our brains, our storms.

Fak­ing work

Our team is instruct­ed to avoid using these tools in any way that’s not explic­it to the cus­tomer. As a client, you should be noti­fied if our team decides, for any rea­son (prag­ma­tism, exper­i­ments, bud­get, dead­lines, etc), that we need to lever­age these tools for faster or bet­ter results.

Fak­ing per­son­al voice

While we can­not and would nev­er con­trol what our team mem­bers say on work-relat­ed social net­works (like LinkedIn, Red­dit, Quo­ra, Twit­ter, Mastodon and sim­i­lar) we do not endorse AI gen­er­at­ed posts on per­son­al accounts.

We strong­ly believe in say­ing things with our own voice.

The jury is still out (we’re the jury) 

This remains an unchart­ed ter­ri­to­ry with more ques­tions than answers. Experts don’t even know how to eval­u­ate LLMs per­for­mance (Azhar & War­ren: Part 1, Part 2 & Part 3). We’re tak­ing adop­tion slow­ly, but ful­ly aware that the tech­nol­o­gy is here to stay.

We set aside a bud­get of hours for our team to play and break things. We have an inter­nal forum where our team mem­bers share their find­ings and dis­cuss pros and cons. We’re ask­ing every­one for input to under­stand the role these tools may have in our dai­ly work. It will con­tin­ue to be a team effort, and a team discussion.

All in all, we’re curi­ous, but not hyped. The gen­er­al feel­ing is we expect­ed Jarvis but got auto­com­plete on its first day on the job.

It’s a brave new world, nonethe­less. (That has such peo­ple in it).

The map is not the territory

AI has been around for quite some time. The new thing is we’ve fed the LLMs mas­sive amounts of data (our data, btw), and it worked well beyond expec­ta­tions. The end prod­uct is an aston­ish­ing sim­u­la­tion of, well, us.

But sim­u­la­tions age bad­ly, or end up becom­ing a par­o­dy. Cre­ation requires expe­ri­ence, it comes from life. Its out­come is the inter­rup­tion of rep­e­ti­tion, prac­ti­cal­ly a miracle.

If you have some­thing to say, don’t let it be ruined by AI’s beau­ti­ful and per­va­sive­ly use­ful imi­ta­tion of cre­ativ­i­ty. Try it, test it, use it, but draw your own lines and pro­ceed with cau­tion. Own your thoughts, cri­tiques, mis­takes, and dis­cov­er­ies. Cel­e­brate and hon­or the unique, irre­place­able, per­sis­tent rel­e­vance of your talent. 

The com­pet­i­tive edge is qual­i­ty, not vol­ume. The real prevails.