Title is an exaggeration, all they found was structures that looked similar to structures in our brains. They have no idea what these structures actually do, and even said there is huge amount of variability in the structures.
As an aside, this:
"They first froze the mini-brains and cut them into ultra-thin sections, which they mounted onto glass slides. They then labeled the sections with different combinations of colored fluorescent tags that are specific to certain cell types, and imaged the sections using an automated scanner."
Shows just how immature the neurobiology field is. I imagine that slicing, and then manually re connecting the slices in a 3d program, must be a pretty painstaking process. Not to mention you are left with no idea how data traveled round the structure. A code analogy would be having a huge codebase handed to you in little chunks, and you have to connect up the pieces by hand. And at the end of it, not even being able to debug it and see if things even work as expected.
This correlation and mapping through ultra thin sections has been recently computerised by 'VAST' and has revealed unexpected nanoscale complexity in neuron connectivity.
I worked in a lab where we had a similar problem reconstructing 3d images from manual sections via the same procedure.
Not only is it a painstaking process, but some slices are lost due to human error while sectioning and the tissue is mechanically distorted when placed onto the slides (squished between the glass and the cover slip).
The solution was simple, albeit much more expensive than sectioning: just MRI scan the organ.
If the sample is small enough you can always use an optical coherence microscope. You can get optical microscope resolutions in 3D of in-vivo samples. It's similar to a confocal microscope but with much higher resolution. Only thing is they can be dang expensive but nowhere near as much as an MRI.
You can also get OCT devices for pretty cheap. They use them in optometrist offices for imaging retinas.
While I like this at a conceptual level, I'm struggling with it on a personal, perhaps moral level. A real brain, trapped, prisoner in a vat... if it's a real brain, it's going to have some level of sentience, just like the rest of us. Yet it is a slave. This is somewhat horrific to me. AI never really bothered me right up until I read this...
> if it's a real brain, it's going to have some level of sentience
Without a sensorimotor apparatus for the brain to manipulate and receive reward-signals from, and thus train to control, there is very likely no "thought" as we would consider it—nothing coherent, no train of thought, no high-level patterns. It's more like the sort of "thought" a foetus would have before its first moment of conscious awareness.
This is a really important point that seems to be overlooked in a lot of discussions around AI/consciousness and such. The key piece is that much of what we consider to be "consciousness" derives from formative experiences of physical reality that aren't available to fetuses. For example, this article[0] discusses whether fetuses can feel physical pain and makes a strong case that in early stages of development (at least until 23 weeks' gestation, if not longer), the physical neural connectivity necessary for the sensation of physical pain are incomplete.
The article also suggests that what we consider to be an experience of physical pain involves certain basic aspects of consciousness that require more advanced postnatal development based on interactions with other people. If this is necessary for physical pain, then more abstract psychological pain would be even further down the road of development. So it's not as simple as whether something is a "real brain" or not, and growing mini-brains from human cells doesn't mean that sentience automatically happens.
I'm inclined to agree that such a mini-brain at this stage is most-likely non-conscious and incapable of suffering, but how would we know?
Organisms with far "simpler" brains than ours appear to be capable of suffering. I'm unaware of the extent to which human embryonic brain development follows expected principles of evolutionary developmental biology, but if that holds, a sufficiently developed brain should at least be given equal ethical consideration as an animal. Vertebrates are commonly born with a capacity for suffering, so I question how much external input is required. If the necessary brain structures are there, those same structures should suffice for achieving brain-states equivalent to percepts of suffering.
Looks to me like we're not quite there yet, as these cerebral organoids are pretty darn small and comparatively disorganized, but it's only a matter of time before the ethical questions need be faced.
Maybe I used a very bad and distracting (and political) example. "Suffering" wasn't really something I was trying to talk about, here. (To be clear, pain is a matter of instinct, which is to say, a predefined brain structure that "works" from birth—and maybe before-hand—whenever it shows up.) Rather, I was specifically talking about this thing we call "thinking." The thing that humans do when they're awake and lucid, and don't do when they're asleep/unconscious/in a dissociative fugue. The thing that the courts care about as mens rea.
Many higher-level vertebrates develop some capacity for "thinking"—that is, for planning/strategizing, puzzle-solving, learning and synthesizing information, forming concepts and abstractions, overriding instincts with beliefs based on evidence, etc. Humans and all other primates do this; as do crows and their relatives; as do dolphins; etc.
But none of these animals start with that ability from birth. "Thinking" has never been observed as behavior available instinctually to any animal. A newborn crow or gorilla, despite being rather self-sufficient with a library of instinctual responses, won't be able to solve the puzzles that we get the adults of those species to solve and which we cite as proof of their intelligence. They grow that ability from months/years of life experience.
In fact, almost by definition, "thinking" is something you have to build: it's a big conceptual hierarchy with sensory data at its base. And—as far as we know—instinct can't pre-bake the needed raw sensory data into brains. (It can bake in higher-level associations, like scent- or pattern-associations for predator avoidance. But it doesn't include these in low-level "raw" terms where a brain can derive any of its own interpretations or abstractions from them.)
Without accumulated experience, a brain is still experiencing the world, and reacting to the world, and experiencing reward signals from the world—like most animals do—but is not yet thinking about the world. Thinking is a matter of updating beliefs about schemas/models/concepts; and those things only exist when derived from sense-data "evidence."
Thanks for sharing your thoughts. I recently had a discussion concerning animal sentience and was having trouble articulating my thoughts. You have managed to concisely convey "thought" in an eloquent fashion.
If this tech evolves, I'm really worried about brain harvesting or trafficking, we already have increasingly desperate people turning to organ trafficking, and human smuggling is tied to organ harvesting.
Unseen suffering is non-existent suffering. With no capacity for understanding what, if anything, another brain is experiencing, it's easy to rationalize and dismiss. That goes for fully formed humans as much as lab experiments.
Every few years, there are incidents of miners trapped underground, with herculean efforts to free the survivors. Each time, I ponder whether there have been similar tragedies, in some corrupt society, where the mine is bulldozed over, any survivors left to die alone in the darkness, none ever knowing of their suffering.
These brains are a lot smaller than a human brain---only about 2-5 mm across. While it's almost certainly a mistake to conflate size with complexity naively (songbirds are capable of pretty complex behaviors), I think it's much safer to assume that if human brains were capable of being much smaller while still allowing complex behaviors, they would be. Otherwise, we would probably gestate longer---compared to lots of other primates, humans are born "premature" in terms of developmental completeness.
Also, as other comments have pointed out, these organoids don't really have any neural inputs, so it's unlikely that they're wired in a way that is at all analogous to real human brains.
This is looking increasingly likely - here's a recent nature paper showing these organoids are electrically active, recording spikes from a dense 256ch electrode. That team also was able stimulate optogenetically yielding bidirectional communication with an organoid - http://www.nature.com/nature/journal/vaop/ncurrent/full/natu...
Another interesting paper demonstrating that neural tissues can be effectively grown around mesh electrodes, providing both a scaffolding function as well a recording and stimulation functions. http://faculty.engr.utexas.edu/xie/xie/publications/ultrafle...
Neuroscience has been rapidly leveling up recently, its increasingly believable that high bandwidth bidirectional neural interfaces (at the cellular level) are on the horizon. DARPA was pitching this a few years ago (NESD) and it was pretty far out, but now see Kernel and Neuralink and the 10 or so companies partnered in to those efforts mostly-successfully developing all kinds of technologies required by this roadmap (disclosure: including my own).
After Elon Musk's Neuralink figures out a scalable way to interface with neurons. Need probably millions of electrodes, not just a few hundred like we use today.
It would be kind of crazy if we develop full AI by literally using brains in vats. But it makes sense (even if it is horrifying). Meat is cheap, and the brain is like an exa-OPS computer running on 20 Watts of power. If you could solve the interface problem and figure out how to actually use it practically, brain is like 6 or 7 orders of magnitude cheaper than the next-cheapest computing substrate. That's like half a century of Moore's Law (and Moore's Law is basically over now... much slower pace, at least).
While dead brains might be cheaper than electronic, I doubt that growing neurons in a lattice and doing what you want will ever be.
There is some research going on in using spintronics for neural network, and I think that both density and power usage would be a fraction of a biological system.
There is probably a long way until we can emulate human brains, but for generic neutral networks, I think electronics or spintronics is the probable route.
Brains are not general purpose computers though. You could probably train lumps of brain tissue do to certain tasks, but putting any kinds of FLOP/s numbers on the label is like stating horse power for levers.
That's why I said "OPS" instead of FLOPS. The kind of operations that brains are good at are orthogonal to the kinds that computers are good at, so a hybrid approach is probably a good idea. Again, I'm primarily thinking about this for AI type applications, not like fluid dynamics or other applications requiring high precision.
EDIT:And technically brains are general computing devices. Just really inefficient ones. Emulating logic in wetware is highly inefficient (and usually requires an external memory device, like pen and paper, though pure wetware memory also works in some especially skilled individuals, although still with severe capacity limitations). Of course, emulating wetware in logic isn't terribly efficient, either.
Some years ago (9+) Professor DeMarse published some papers on training petri dish brains cells how to "fly a plane" (or more appropriately "keep a simulated plane from crashing"). I have a copy of the paper at home, but can't think of the title, (it was kind of hard to find a copy since it wasn't publicly available online). You might look at what his lab is currently working on though.
As an aside, this:
"They first froze the mini-brains and cut them into ultra-thin sections, which they mounted onto glass slides. They then labeled the sections with different combinations of colored fluorescent tags that are specific to certain cell types, and imaged the sections using an automated scanner."
Shows just how immature the neurobiology field is. I imagine that slicing, and then manually re connecting the slices in a 3d program, must be a pretty painstaking process. Not to mention you are left with no idea how data traveled round the structure. A code analogy would be having a huge codebase handed to you in little chunks, and you have to connect up the pieces by hand. And at the end of it, not even being able to debug it and see if things even work as expected.