Arti­fi­cial Intel­li­gence (AI) and Ethics

Tech­nol­o­gy is com­ing clos­er and clos­er to the essen­tial and per­son­al affairs of human beings. Whether it be biotech­nol­o­gy, med­i­cine or dig­i­tal appli­ca­tions and apps that record and impact our dai­ly inter­ac­tions, they all influ­ence humans as a col­lec­tive as well as indi­vid­u­als when deci­sions are made. The call for ethics is grow­ing loud­er and loud­er. The pan­dem­ic has led to increased eth­i­cal debates sur­round­ing the big ques­tions of if, and, if so, which eth­i­cal under­stand­ing will guide human­i­ty, how ethics will affect future devel­op­ments in the dig­i­tal age and how eth­i­cal val­ues can and should be respected. 

In the lat­est of her Duet inter­views, Dr Cal­daro­la, author of Big Data and Law, and AI expert Leila Taghizadeh dis­cuss the com­pat­i­bil­i­ty of ethics and AI.

To start off our con­ver­sa­tion, can you first give our read­ers a short intro­duc­tion to what AI (Arti­fi­cial Intel­li­gence) is and how it works?

Dr Leila Taghizadeh: Basi­cal­ly, it is the intel­li­gence shown by a machine or any device built by humans. You might already con­sid­er all com­put­ers to be intel­li­gent because some of them can per­form tasks which are way beyond the abil­i­ties of even the most intel­li­gent peo­ple; a sim­ple exam­ple being the appli­ca­tions which can solve com­plex math equa­tions. How­ev­er, we are look­ing at a slight­ly dif­fer­ent sit­u­a­tion: In the world of “tra­di­tion­al” com­put­ers, a human cre­ates the algo­rithm and the pro­gram while the com­put­er only fol­lows cer­tain com­mands and exe­cutes exact­ly what it has been told to do! Imag­ine a gate where you could drop off your parcels to be processed. With a tra­di­tion­al com­put­er, you would have to define dif­fer­ent types of parcels and make them known to the machine. Oth­er­wise, the com­put­er wouldn’t know what to do when it received a box instead of an enve­lope, for exam­ple. So, in this hypo­thet­i­cal pro­gram, you define that a par­cel can be an enve­lope or a box. Now, if you try to drop off a bag, the machine wouldn’t know what to do, and, depend­ing on how it was pro­grammed, either it would sim­ply do noth­ing or it would give an error mes­sage. In the world of AI, how­ev­er, a machine looks for sim­i­lar­i­ties between a bag and a box and an enve­lope and will treat it as a par­cel. A machine is capa­ble of learning!

Is AI always dri­ven by a human being – in oth­er words, will there always be a per­son at the end of a devel­op­ment– in your arti­cle you refer to them as AI tech­nol­o­gists – who can be traced back to the out­come of AI-dri­ven tech­nol­o­gy and can, there­fore, be held account­able? Or will AI also become an inde­pen­dent result of machine learn­ing with­out the influ­ence, impe­tus or guid­ance of a human being because machines/robots will be able to improve and advance them­selves? If so, who is liable-espe­cial­ly if the cur­rent laws only attribute prof­its, loss and dam­ages to human beings?

This is a hot top­ic in the domain of AI; the reg­u­la­tions guid­ing the devel­op­ment of AI still have to catch up with the devel­op­ment of AI. The nature of AI should enable machines to oper­ate with­out human guid­ance or inter­fer­ence at some lev­el- at least at some point in the future. How­ev­er, humans will prob­a­bly still be at the begin­ning of the chain:  at that moment in time when the AI in ques­tion is being cre­at­ed. That is why we need to make sure that the rel­e­vant algo­rithms are with­out bias or dis­crim­i­na­tion. You can­not build an AI based on a dis­crim­i­na­tive algo­rithm and expect it to behave “fair­ly”. Now, here comes the prob­lem: We, as humans, have lots of con­scious and uncon­scious bias­es; we cre­ate rules in our soci­ety based on our bias­es and we rule soci­ety with our bias­es. So how can we expect AI not to have them? Although with AI, we have a big­ger prob­lem: The bias­es will be prop­a­gat­ed much faster and in big­ger dimen­sions! Can we change this? I still believe we can, but it will need to be a con­scious deci­sion and a col­lec­tive effort.

I see anoth­er inter­est­ing angle to your ques­tion which con­cerns lia­bil­i­ty. When we cre­at­ed cor­po­ra­tions and insti­tu­tions and gave them pow­er of attor­ney, how did we answer this ques­tion of lia­bil­i­ty? A regime, a gov­ern­ment, a cor­po­ra­tion has a cer­tain pow­er over human life at var­i­ous lev­els– who are these insti­tu­tions? Are they human? Or “arti­fi­cial or imag­i­nary” objects?  The world of AI won’t be dif­fer­ent in con­cep­tu­al terms from today’s institutions.

By def­i­n­i­tion, AI includes the word “intel­li­gence”. As we know, AI works with quan­ti­ta­tive meth­ods and cor­re­la­tions and results of algo­rithm which lead to “the biggest, the most fre­quent etc”. Can this kind of tech­nol­o­gy be “intel­li­gent”?

I would say it depends on how we define intel­li­gent. We have dif­fer­ent types of AI¹:

• Nar­row or Weak AI: This is what is main­ly used today; e.g., for facial recog­ni­tion, speech recognition/voice assis­tants, dri­ving a car, or search­ing the inter­net – and this tech­nol­o­gy is very intel­li­gent at com­plet­ing the spe­cif­ic task for which it is pro­grammed to do.
• Strong or Deep AI: This tech­nol­o­gy has not yet been achieved; this is a type of AI for which we need to find a way to make machines con­scious. Machines would have to take expe­ri­en­tial learn­ing to the next lev­el, not just improv­ing effi­cien­cy of indi­vid­ual tasks, but gain­ing the abil­i­ty to apply knowl­edge gained from expe­ri­ence to a wider range of dif­fer­ent prob­lems. This sort of AI can be achieved in the near future as well; All we need is for a com­put­er to be exposed to many dif­fer­ent sit­u­a­tions and learn from them. It might not be eas­i­ly attain­able for all domains, but we have made good progress regard­ing this type of AI in the labs in sit­u­a­tions where AI can recog­nise human feel­ings or their tone and adapt their respons­es accordingly.
• The Super­in­tel­li­gence AI is a type of AI where machines become self-aware and sur­pass the capac­i­ty of human intel­li­gence and abil­i­ty. This type only exists in the world of sci-fi – at least so far; but in my opin­ion, even this type can even­tu­al­ly be achieved. After all, through­out his­to­ry, we humans have cre­at­ed every­thing we dreamed of.

You work in the domain of AI and ethics and you were mem­ber of a group that devel­oped eth­i­cal prin­ci­ples for AI. Are those eth­i­cal prin­ci­ples meant to be a gen­er­al guide­line for AI tech­nolo­gies or machines? Or are they intend­ed to be the basis for an ISO stan­dard or an inter­na­tion­al treaty? Will or should they be part of a company’s cor­po­rate respon­si­bil­i­ty, cor­po­rate gov­er­nance or code of con­duct? Or do they rep­re­sent new inno­va­tion goals as a renun­ci­a­tion of Six Sig­ma? How will this trans­late in con­crete terms: Will human beings learn and abide by the eth­i­cal prin­ci­ples or will ethics be a tech­ni­cal part of soft­ware that influ­ence the AI process (ethics by design)?

Well, in an ide­al world, ethics should be part of the struc­ture! This is the only way we can make com­plete use of AI in an eth­i­cal way. How­ev­er, ethics vary from nation to nation. This cre­ates some con­cern and no straight­for­ward solu­tion has yet to be found.  But will it hap­pen at a broad lev­el? I am not sure. My hope, and the effort of the com­mu­ni­ty, has been to ensure that we have some stan­dards which can be altered or adjust­ed. After all, we have rules in any soci­ety so why not in this case? The sit­u­a­tion, how­ev­er, is a bit more com­plex: How well can we assess just how fair and eth­i­cal these ele­ments are? Should we destroy an AI if the algo­rithm is not fair? Or would we be unfair­ly pun­ish­ing the cre­ator? We might want to bor­row some con­cepts from our cur­rent mod­els for com­pa­nies: When a com­pa­ny com­mits a crime on a big scale, what are the consequences?

I realise that I am ask­ing more ques­tions than pro­vid­ing answers, but I think this is the role of the com­mu­ni­ty: to ask ques­tions, to cre­ate aware­ness regard­ing unre­solved issues, and to make those who come up with the rules more accountable.

As far as I am aware, machines do not pos­sess innate­ly human qual­i­ties nor emo­tions although sci­en­tists pre­dict that it is only a mat­ter of time before they are capa­ble of this as well. Will the quan­tum com­put­er become a game chang­er? Will machines devel­op emo­tions or only the result of what they have learnt will have emo­tion­al, eth­i­cal and human traits?

For Strong AI, quan­tum com­put­ing would be of great assis­tance. We sim­ply need to increase the speed of the test data being gen­er­at­ed and of the algo­rithm being exposed to that data:  The faster we achieve this sit­u­a­tion in tech­no­log­i­cal terms, the soon­er we arrive at the results you are talk­ing about. More­over, we can­not for­get that for cer­tain com­plex tasks, today’s com­put­ers can­not deliv­er the required power.

We also have to con­sid­er anoth­er mat­ter:  We not only need quan­tum com­put­ing to be able to have more pow­er, but also to use less ener­gy. Look­ing at today’s ener­gy con­sump­tion and glob­al warm­ing, I think we can all agree that we won’t be able to live on this plan­et, if we extrap­o­late the required ener­gy for com­put­ing need­ed for a con­comi­tant increase in com­put­er pow­er. There­fore, we need to reduce the ener­gy required. Such a con­di­tion can be achieved with­in the domain of Nano quan­tum com­put­ing, where, sim­ply stat­ed, elec­trons gen­er­ate cur­rent with­out con­sum­ing any ener­gy but only by their Brown­ian move­ment. That is anoth­er excit­ing top­ic in this domain.

There is a com­mon under­stand­ing that a busi­ness person’s pri­ma­ry moti­va­tion and goal is to increase the prof­its of his or her busi­ness and that s/he will prob­a­bly not care about noble (eth­i­cal) moti­va­tions. The same could be said of tech­nol­o­gists, isn’t that so? Are they also only inter­est­ed in fos­ter­ing inno­va­tion- regard­less of laws, ethics and fields of appli­ca­tions? Or, in oth­er words, will and can a tech­nol­o­gist cease to sat­is­fy his or her curios­i­ty because the out­come might have a neg­a­tive impact? Is the ori­en­ta­tion of a tech­nol­o­gist only towards fea­si­bil­i­ty and pro­ducibil­i­ty instead of eth­i­cal duties?

I think this might be true of some peo­ple.  How­ev­er, we shouldn’t gen­er­alise. Regard­less of the top­ic, whether it be AI or some­thing else, there are smart respon­si­ble peo­ple as well as some nar­cis­sists among us who are equal­ly intel­li­gent. Some will cre­ate tech­nolo­gies to improve the qual­i­ty of life for oth­ers, while oth­ers might only cre­ate tech­nolo­gies for their own ben­e­fit! We have seen dif­fer­ent appli­ca­tions of dig­i­tal tech­nolo­gies, and I believe it will be the same for AI! As in every domain, every­one (fair or self­ish, eth­i­cal or non-eth­i­cal) will be able to par­tic­i­pate in this domain. This means that pre­sum­ably reg­u­lat­ing this domain will become nec­es­sary- at least in the long term.

The dig­i­tal world has made the world more acces­si­ble and less depen­dent on the nobil­i­ty. It has resolved lots of prob­lems, but, at the same time, it has cre­at­ed plen­ty of new ones. In the world of digi­ti­sa­tion, we could eas­i­ly lis­ten to lots of moti­va­tion­al TED talks, have access to edu­ca­tion, become more aware and sup­port­ive of peace­ful move­ments. At the same time, racist and crim­i­nal ele­ments could also seek more sup­port and expand more eas­i­ly and faster. How­ev­er, if we ask our­selves what has been the net effect of the dig­i­tal world, I think we would all agree that it has been, on the whole, more pos­i­tive. This devel­op­ment is also the next step to be tak­en by our civil­i­sa­tion; it is a one-way street and there is no way back. We have grown up in this world, devel­oped skills which are required to live in this world, and, in few hun­dred years, human intel­li­gence and brains will be com­plete­ly dif­fer­ent from a few hun­dred years ago.

When look­ing at life on a time­line, any dis­cov­ery or step is just a step; it will have a past and a future, and the rise of AI will be the same.

My opin­ion is:

“AI is just anoth­er step in human development!”

Dr Leila Taghizadeh, MBA

One of the best exam­ples is the mag­ni­tude of dif­fer­ent pri­va­cy laws – we need only look at the Euro­pean Com­mu­ni­ty, the US and Chi­na. In order to come to a com­mon under­stand­ing, a com­mon set of ethics– at least in demo­c­ra­t­ic states – is nec­es­sary. How can peo­ple agree on a com­mon set of ethics if they do not have a shared his­to­ry that leads them to the same/similar eth­i­cal per­cep­tions? Or will machines and tech­nol­o­gy solve this prob­lem because they will apply the same eth­i­cal rules so that machines and tech­nol­o­gy will teach us what is or is sup­posed to be ethical?

I do not think we will reach a com­mon agree­ment; we will still grow region­al­ly and will have some inter­na­tion­al umbrel­la rules; oth­er reg­u­la­tions have devel­oped in this way, after all. This might pro­hib­it suing a spe­cif­ic AI from one region in anoth­er. I think AI – at least in the short term – will not change the nature of our civil­i­sa­tion. With the advent of the Super Intel­li­gence AI, we might see the next rev­o­lu­tion (after the agri­cul­tur­al and indus­tri­al rev­o­lu­tions). How­ev­er, I do hope that we have a more acces­si­ble and fair­er world by then!

Leila, thank you for shar­ing your insights on the com­pat­i­bil­i­ty of ethics and AI.

Thank you, Cristi­na, and I look for­ward to read­ing your upcom­ing inter­views with recog­nised experts, delv­ing even deep­er into this fas­ci­nat­ing topic.

---

¹ https://​www​.investo​pe​dia​.com/​t​e​r​m​s​/​w​/​w​e​a​k​-​a​i​.​asp