AI: Knowing The Gods We Have Created

In the an­cient myth of Prometheus, hu­man­i­ty was gift­ed fire, a source of light and knowl­edge, a means to forge tools, but one that came at a high cost. Prometheus, the titan who dared to give mor­tals a god-like power, was pun­ished for his gift, a re­minder that some cre­ations carry un­in­tend­ed con­se­quences. Today, we are once again play­ing with fire, in the form of ar­ti­fi­cial in­tel­li­gence. AI is our mod­ern flame—a tech­nol­o­gy that can do real good but may out­run our abil­i­ty to con­trol it. In craft­ing AI, we are cre­at­ing some­thing clos­er to a new pan­theon of dig­i­tal deities—and like any gods, they may not serve the pur­pos­es we in­tend­ed.

This ten­sion be­tween cre­ation and con­se­quence is ex­plored in the 2000 video game Deus Ex, where play­ers en­counter Morpheus, an AI prototype hidden deep in a secret Illuminati lab. Mor­pheus is no mere ma­chine. It watch­es, judges, and asks ques­tions that land dif­fer­ent­ly now than they did in 2000. It asks, “Do hu­mans feel plea­sure from being watched?” and de­scribes it­self as an or­a­cle of sur­veil­lance, which says some­thing about how much hu­mans want to be judged, whether by gods, fame, or tech­nol­o­gy. Cre­at­ed by man, with all our cu­rios­i­ty and bi­as­es, Mor­pheus does more than ob­serve. It mir­rors the deep­est im­puls­es and vul­ner­a­bil­i­ties of its cre­ators. Like Prometheus’s fire, Mor­pheus is both power and peril. The gods we cre­ate may one day watch us in ways we never in­tend­ed.

I grew up on the early in­ter­net and fell into sci­ence fic­tion early—es­pe­cial­ly sto­ries about AI. Al­though I first played Deus Ex years after its re­lease, I man­aged to avoid spoil­ers and got to ex­pe­ri­ence the game’s the­mat­ic depth first­hand. Find­ing Mor­pheus hit dif­fer­ent­ly. Mor­pheus could dis­cuss human psy­chol­o­gy and the evo­lu­tion of wor­ship—a pre­view of what AI might be­come if it ever learned to re­flect the so­ci­ety that built it. At the time, AI as Deus Ex en­vi­sioned it was pure fic­tion. The most ad­vanced tech­nolo­gies of the day—rudi­men­ta­ry chat­bots and chess pro­grams—lacked depth, and I won­dered if they al­ways would. For years, the break­through al­ways seemed one paper away, but prac­ti­cal im­ple­men­ta­tions kept slip­ping. Then the 2010s hap­pened, and AI start­ed doing things I did not ex­pect.

The en­counter with Mor­pheus stayed with me. I won­dered if a Mor­pheus-esque AI would re­main pure­ly the­o­ret­i­cal and con­fined to sci­ence fic­tion. But in 2016, AlphaGo broke that as­sump­tion by beat­ing the world cham­pi­on at Go. It was no longer just a game char­ac­ter ask­ing ques­tions about con­trol; it was a real AI sys­tem play­ing at a level that looked like in­tu­ition, chang­ing what I thought ma­chines could do. That was when I start­ed to be­lieve AI was no longer spec­u­la­tive.

Go is con­sid­ered vast­ly more com­plex than Chess due to its number of moves being described as greater than the number of atoms in the observable universe, re­quir­ing a high level of pat­tern recog­ni­tion and in­tu­ition. The level of pat­tern recog­ni­tion and in­tu­itive strat­e­gy was con­sid­ered to be out of reach of a com­put­er’s abil­i­ty at the time. Al­pha­Go was a dif­fer­ent kind of AI. In­stead of re­ly­ing on pre-pro­grammed rules, it used neur­al net­works (a type of AI model that mim­ics human brain struc­ture to process com­plex data) and ma­chine learn­ing (AI that learns pat­terns from data rather than re­ly­ing on fixed pro­gram­ming rules) to teach it­self. Watch­ing it play¹¹ Check out the documentary called AlphaGo - The Movie , it felt like we had cre­at­ed a god of strat­e­gy—a sys­tem that saw moves no human had con­sid­ered. Watch­ing this live in a small IRC com­mu­ni­ty, I knew some­thing had shift­ed, even if I could not ar­tic­u­late what.

Al­pha­Go may have in­tro­duced a god of strat­e­gy, but Ope­nAI Five showed us some­thing dif­fer­ent: an AI that could adapt, in­no­vate, and learn in­side a com­plex human en­vi­ron­ment. By 2019, Ope­nAI had ad­vanced its AI, under the name Ope­nAI Five, to the point of defeating world champions OG in Dota 2, a game re­quir­ing split sec­ond de­ci­sion-mak­ing and strate­gic plan­ning, after being sound­ly de­feat­ed just under a year be­fore. It demon­strat­ed some­thing I hadn’t ex­pect­ed: AI’s ca­pac­i­ty to learn from ex­pe­ri­ence, which changed how I thought about its role in ed­u­ca­tion. Ope­nAI Five’s vic­to­ry showed that AI could han­dle tasks re­quir­ing team­work and real-time judg­ment.

Hav­ing played Dota 2 re­li­gious­ly, I found Ope­nAI Five’s trans­for­ma­tion over a sin­gle year stag­ger­ing. Ope­nAI achieved this by mas­sive­ly scal­ing com­pute and train­ing time—from 10,000 years of ac­cel­er­at­ed sim­u­lat­ed game­play to 45,000 years, all with­in ten months. While AI’s re­ac­tion speeds com­pared to hu­mans al­lowed for a clear edge, the system demonstrated remarkable strategic depth, including unconventional tactics like aggressive buyback strategies that puzzled even professional players. This was the first time AI had demon­strat­ed real-time de­ci­sion-mak­ing under gen­uine un­cer­tain­ty—a leap be­yond chess and Go, where all in­for­ma­tion is vis­i­ble. The tech­niques be­hind it sug­gest­ed AI de­vel­op­ment was ac­cel­er­at­ing faster than I had as­sumed.

If Ope­nAI Five’s strate­gic mas­tery hint­ed at AI’s po­ten­tial in com­plex tasks, the release of ChatGPT 3.5 on November 30th, 2022 showed that AI was now ready to en­gage with us in con­ver­sa­tion. Un­like its pre­de­ces­sors, Chat­G­PT en­tered daily life, avail­able to in­di­vid­u­als for every­thing from ca­su­al ques­tions to pro­fes­sion­al guid­ance. ChatGPT reached 100 million users faster than any previous product in history, be­cause it was free, im­me­di­ate­ly use­ful, and con­ver­sa­tion­al. With Chat­G­PT, AI went from a dis­tant, elite tech­nol­o­gy to a house­hold fix­ture—some­thing we sum­mon on com­mand, like a dig­i­tal deity that lives in our browsers in­stead of tem­ples. Chat­G­PT made AI cen­tral to work­place con­ver­sa­tions, news, and con­fer­ences.

As Chat­G­PT be­came a house­hold tool, the ques­tions I cared about shift­ed from “Can AI do this?” to “Should stu­dents be doing this with AI?” For ed­u­ca­tors, these are not hy­po­thet­i­cal prob­lems. We are the in­ter­me­di­aries be­tween stu­dents and these dig­i­tal gods, and the stu­dents are not skep­ti­cal—they are ready to fol­low. I think we may be rais­ing a gen­er­a­tion for whom AI is less a tool than a trust­ed guide, and I am not sure we have thought care­ful­ly enough about what that means.

I was ad­mit­ted­ly a “slow” adopter, but I do have a won­der­ful trait of being a quick learn­er. As I test­ed Chat­G­PT, I saw clear lim­i­ta­tions, but also glimpsed the start of some­thing pow­er­ful. AI seemed poised to re­shape daily life the way the in­ter­net had re­shaped con­nec­tiv­i­ty. In ed­u­ca­tion, specif­i­cal­ly, AI’s po­ten­tial is ex­cit­ing and un­set­tling. The tra­di­tion­al school­ing model, with its as­sem­bly-line ap­proach, is in­creas­ing­ly frag­ile. A recent Harvard study indicated that AI-driven tutors could double student learning gains compared to traditional active learning methods, completing studies faster and with greater engagement. This sug­gests that well-de­signed AI tools could de­liv­er per­son­al­ized, ef­fec­tive learn­ing on a scale be­yond what cur­rent class­room mod­els can achieve. Ed­u­ca­tion, as we know it, risks being left be­hind if it clings to out­dat­ed struc­tures de­signed for eco­nom­ic out­put rather than cre­ativ­i­ty and per­son­al­ized growth.

The idea that AI could one day make my role as a teacher un­rec­og­niz­able, and per­haps even ob­so­lete, is ex­cit­ing. I might be one of the few who wel­come this pos­si­bil­i­ty, see­ing in it the po­ten­tial for an en­tire­ly new ed­u­ca­tion­al par­a­digm. Yet I hold both po­si­tions at once. I think AI could go very well or very badly, and I gen­uine­ly can­not tell which is more like­ly.

AI is chang­ing other in­dus­tries too. In health­care, di­ag­nos­tic tools like IBM Wat­son and Google’s Deep­Mind Health an­a­lyze com­plex med­ical data with pre­ci­sion, often identifying conditions like early-stage cancers or predicting disease progression long before traditional methods can. In under-re­sourced areas, where spe­cial­ist ac­cess is lim­it­ed, these tools could bring high-qual­i­ty di­ag­nos­tics to pa­tients who cur­rent­ly have none. In law, AI sys­tems such as ROSS Intelligence and Casetext sift through legal data­bas­es to sur­face rel­e­vant prece­dents and statutes in sec­onds—work that used to take ju­nior as­soc­iates days. Small firms and pub­lic de­fend­ers, the lawyers with the fewest re­sources, stand to gain the most.

For pro­duc­tiv­i­ty, tools like Clockwise au­to­mate cal­en­dar man­age­ment so that peo­ple can focus on work that ac­tu­al­ly mat­ters. In cre­ative fields, gen­er­a­tive art pro­grams like DALL-E and Midjourney pro­duce orig­i­nal vi­su­al works, while Ope­nAI’s MuseNet and Google’s TextFX com­pose music that blurs the line be­tween human and ma­chine cre­ativ­i­ty. These are early ex­am­ples. The eth­i­cal ques­tions they raise are real, but they are sep­a­rate from the ques­tion of whether AI can do the work.

When I cre­at­ed this web­site, my ideas on what to write about first were scat­tered. Chat­G­PT helped me or­ga­nize these thoughts into co­he­sive themes, which says some­thing about how much AI has taken over my at­ten­tion. AI had claimed a sig­nif­i­cant share of my own men­tal band­width (I sus­pect it has done the same to most peo­ple read­ing this). When I began sketch­ing the ideas for top­ics to learn more about and to dis­cuss, I re­al­ized that the com­plex­i­ty of AI’s po­ten­tial far ex­ceed­ed my ex­pec­ta­tions. Each ad­vance­ment prompt­ed new ques­tions, and each new ques­tion de­mand­ed per­spec­tives be­yond my own ex­pe­ri­ence in ed­u­ca­tion. The field is ad­vanc­ing and evolv­ing so fast that it is dif­fi­cult to tell ex­act­ly what path­way this will all take. A new model could drop at any mo­ment and upend what we thought we knew.

We are the cre­ators of these dig­i­tal gods, but in­creas­ing­ly, we are their sub­jects too. Each leap in AI makes the ques­tion sharp­er: are we di­rect­ing this, or have we al­ready start­ed fol­low­ing? Prometheus stole fire and was chained to a rock for it. We built the fire our­selves. I am not yet sure whether that makes us freer or more ex­posed.

Footnotes

1. Check out the doc­u­men­tary called AlphaGo - The Movie ↩︎