Amoebas are smarter than they look, and a team of US physicists think they know why. The group has built a simple electronic circuit that is capable of the same 鈥渋ntelligent鈥� behaviour as Physarum, a unicellular organism 鈥� and say this could help us understand the origins of primitive intelligence.
In recent years, the humble amoeba has surprised researchers with its ability to behave in an 鈥渋ntelligent鈥� way. Last year, Liang Li and Edward Cox at Princeton University reported that the Dictyostelium amoeba is twice as likely to turn left if its last turn was to the right and vice versa, which suggests the cells have a rudimentary memory.
This year, at Hokkaido University in Sapporo, Japan, for his work on amoeba intelligence after his team found further evidence of the amoeba memory effect. They exposed Physarum amoeba to temperatures fluctuating regularly between cold and warm. It was already known that the cells become sluggish during cold snaps, but Nakagaki鈥檚 team found that the amoeba slowed down in anticipation of cold conditions, even when the temperature changes had stopped ().
Advertisement
, and Steven La Fontaine at the University of California, San Diego, wondered what was responsible for that behaviour.
In the past, biologists have suggested that there are natural oscillators within the cells that can change their frequency in response to a changing environment. But that can鈥檛 be the complete picture, say the researchers, because the amoeba鈥檚 response is short-lived. For instance, although Physarum 鈥渓earned鈥� to anticipate cold spikes in Nakagaki鈥檚 experiments, it quickly 鈥渞ealised鈥� that those cold spikes were no longer occurring and stopped anticipating them.
Instead, Di Ventra鈥檚 team thinks there is an intrinsic memory storage device within the amoeba. As with the human brain, that device can strengthen and store memories for some time. But if the memory isn鈥檛 used, it gradually fades away.
Liquid memory
Now they have identified a potential storage device. The amoeba鈥檚 interior contains a watery sol 鈥� a solid suspended in liquid 鈥� within a thick viscous gel. The sol flows through the gel like water through a sponge, creating a network of low-viscosity channels. Those channels are strengthened as long as the amoeba continues to respond to a static environment, but if that environment changes the channels gradually break down and a new network appears as the amoeba adapts. For a short while, though, the amoeba retains a 鈥渕emory鈥� of those earlier conditions.
Di Ventra鈥檚 team took advantage of the development this year of memristors 鈥� electrical resistors that retain a memory of earlier voltages or currents applied and vary their resistance accordingly 鈥� to design a simple circuit that models the amoeba鈥檚 gel-sol system. Their circuit contained just four basic elements: a resistor, capacitor, inductor and memristor. By changing the external voltage in a regular way they could model the changing temperature conditions studied by Nakagaki鈥檚 team. When they did this, they found that their circuit could 鈥渓earn鈥� and predict future voltage fluctuations.
鈥淚t appears that our model describes pretty well the experiments on amoebas鈥� learning,鈥� says Di Ventra. He cautions there is a huge gap between the simple response of single-celled animals and the cognitive abilities of developed species, but adds there is no doubt that a combined set of simple circuit models will have more complex behaviour. 鈥淭his is in fact what we are now interested in studying,鈥� he says.
But others remain unconvinced that the behaviour of amoebas can be explained by circuit models. 鈥淚n my view, simulation is not an adequate approach for such a deep biological problem as intelligence,鈥� says Guenter Albrecht-Buehler at Northwestern University Medical School in Chicago. 鈥淲e need to find out how it鈥檚 really done in a living cell.鈥� He points out that all sorts of electronic models could have simulated genetic memory but wouldn鈥檛 have helped to understand how cells achieve that particular trick. 鈥淔inding the double helix of DNA was the only achievement that really mattered,鈥� he says.
Klaas Hellingwerf at the University of Amsterdam in the Netherlands questions the Di Ventra team鈥檚 definition of the amoeba鈥檚 behaviour. 鈥淚 have reservations in calling this 聭learning鈥�, let alone 聭intelligence鈥�,鈥� he says. Detecting and anticipating regular signals doesn鈥檛 imply learning, he explains. 鈥淚 think learning implies more 鈥� for instance, associative memory.鈥� An example of associative memory would be if the cell learned a response to one stimulus and could then use that response when faced with a second stimulus that is similar to the first.
A pre-print of the Di Ventra paper is available at .
