Jensen Huang’s errings were all too human – so were his triumphs
AI chatbots may not have extramarital affairs, but that doesn’t make their ‘judgement’ any sounder.
Want unlimited, ad-free access? Become a spiked supporter.
Jensen Huang, the man who turned Nvidia into the indispensable hardware supplier for the AI boom, has resigned after his extramarital affairs were revealed to the world. These include a relationship with a woman whom US officials reportedly suspect of links to Chinese intelligence and another with a senior Nvidia employee. There are serious questions about conflicts of interest, personal failings, corporate governance and national security. Huang’s judgement can quite reasonably be called into question.
Apparently, however, Huang is not the only defendant in the dock. Humanity is there with him, too.
In her Financial Times column, Emma Jacobs uses Huang’s downfall to question whether human judgement is really ‘all it’s cracked up to be’. She uses an individual’s mistakes to indict the species. Here, Jacobs is making a category error. The problem with Huang does not lie in human judgement in itself. It rests on the fact that he exercised it badly.
Human judgement obviously isn’t infallible. We can exercise it brilliantly, carelessly or disgracefully. More to the point, we can use our own judgement to judge how another has exercised theirs, as Jacobs herself has done in the case of Huang’s affairs.
You might not want Huang to marry your daughter, but you sure as hell would want him building your chip company. He and his colleagues have helped transform computing, not least through CUDA, which enabled researchers to harness Nvidia’s graphics processors for work far beyond graphics.
That achievement is not diminished by his private misconduct. Nor does it exempt Huang from scrutiny. Different judgements are required when assessing an individual’s private conduct and public achievements.
Steve Jobs is a case in point. According to his eldest daughter, Lisa Brennan-Jobs, he refused to acknowledge her as his daughter until she was nine and maintained a distant and cold relationship with her thereafter. Her account makes for painful reading. But none of this renders his achievements, from the Macintosh personal computer to the iPhone, any less remarkable.
We should resist the demand for a single all-encompassing verdict on a person, whether that means celebrating him regardless of his failings or damning him and everything he accomplished. It is important instead to recognise that someone can possess extraordinary abilities and serious flaws.
The deeper paradox concerns AI itself. Some suggest that AI is making human judgement redundant. But this is to misunderstand both AI and human judgement. After all, who designs the AI model, selects its training data, establishes its objectives and decides which outputs are acceptable? Who determines whether its performance justifies putting it into use? Humans, of course.
These are not incidental contributions to an otherwise autonomous achievement. They are the foundation on which these achievements rest. In OpenAI’s account of how it trains its models to follow instructions, human researchers demonstrate desirable responses and rank competing outputs. Those assessments are then used to shape the model’s behaviour. The researchers explicitly acknowledge that the problems being addressed cannot be fully captured by simple automatic measures. Human judgement is not an embarrassing residue in AI’s development. It is precisely how an unconscious computing programme is made useful.
This does not mean every undesirable output can be eliminated by more careful programming. It means that distinguishing a worthless achievement from a worthwhile one requires judgement. A machine can outperform us in a particular task. But it’s humans who decide which tasks matter and which tasks don’t.
In the Irish Times, political psychologist Joris Lammers talks of how accountability can moderate the behaviour of powerful people. And surely this is the point? Fallible human beings need other human beings to question them, challenge their excuses and hold them to account. That is an argument against treating a successful executive as an oracle.
Huang’s disgrace does not discredit the human capacity to judge; it merely shows why we need to exercise it more rigorously, especially when wealth, power and past achievement tempt us to suspend it. Our fallibility demands our vigilance. We need to question decisions, confront harmful consequences and hold those responsible for failures of judgement to account. We do not overcome our capacity for error by surrendering the very faculty through which we recognise and correct it. Nor can we escape that responsibility by delegating decisions to machines whose design and use depend on human choices.
The answer to bad judgement is not the abdication of judgement. It is the courage to exercise it and the willingness to answer for it.
Dr Norman Lewis is a senior research fellow at the Freedom and Sovereignty Initiative Europe (FSIE). His Substack is What a Piece of Work is Man!
You’ve hit your monthly free article limit.
Support spiked and get unlimited access.
Support spiked and get unlimited access
spiked is funded by readers like you. Only 0.1% of regular readers currently support us. If just 1% did, we could grow our team and step up the fight for free speech and democracy.
Become a spiked supporter and enjoy unlimited, ad-free access, bonus content and exclusive events – while helping to keep independent journalism alive.
Monthly support makes the biggest difference. Thank you.
Comments
Want to join the conversation?
Only spiked supporters and patrons, who donate regularly to us, can comment on our articles.